Virginia Tech to Build Top 5 Supercomputer?

hype7 writes "ThinkSecret is running a story which might explain exactly why the Dual 2GHz G5 machines have been delayed to the customers that ordered them minutes after the keynote was delivered. Apparently, Virginia Tech has plans to build a G5 cluster of 1100 units. If it manages to complete the cluster before the cut-off date, it will score a Top 5 rank in the Linpack Top 500 Supercomputer List. Both Apple and the University are playing mum on the issue, but there's talk of it all over the campus."

Comments

GNAA!

[Anonymous Coward]

GNAA (GAY NIGGER ASSOCIATION OF AMERICA) is the first organization which
gathers GAY NIGGERS from all over America and abroad for one common goal - being GAY NIGGERS.

Are you GAY ?
Are you a NIGGER ?
Are you a GAY NIGGER ?

If you answered "Yes" to any of the above questions, then GNAA (GAY NIGGER ASSOCIATION OF AMERICA) might be exactly what you've been looking for!
Join GNAA (GAY NIGGER ASSOCIATION OF AMERICA) today, and enjoy all the benefits of being a full-time GNAA member.
GNAA (GAY NIGGER ASSOCIATION OF AMERICA) is the fastest-growing GAY NIGGER community with THOUSANDS of members all over United States of America. You, too, can be a part of GNAA if you join today!

Why not? It's quick and easy - only 3 simple steps!

First, you have to obtain a copy of GAY NIGGERS FROM OUTER SPACE THE MOVIE and watch it.

Second, you need to succeed in posting a GNAA "first post" on slashdot.org, a popular "news for trolls" website

Third, you need to join the official GNAA irc channel #GNAA on EFNet, and apply for membership.
Talk to one of the ops or any of the other members in the channel to sign up today!

If you are having trouble locating #GNAA, the official GAY NIGGER ASSOCIATION OF AMERICA irc channel, you might be on a wrong irc network. The correct network is EFNet, and you can connect to irc.secsup.org or irc.isprime.com as one of the EFNet servers.
If you do not have an IRC client handy, you are free to use the GNAA Java IRC client by clicking here.

If you have mod points and would like to support GNAA, please moderate this post up.

This post brought to you by Penisbird , a proud member of the GNAA

G_____________________________________naann_______ ________G
N_____________________________nnnaa__nanaaa_______ ________A
A____________________aanana__nannaa_nna_an________ ________Y
A_____________annna_nnnnnan_aan_aa__na__aa________ ________*
G____________nnaana_nnn__nn_aa__nn__na_anaann_MERI CA______N
N___________ana__nn_an___an_aa_anaaannnanaa_______ ________I
A___________aa__ana_nn___nn_nnnnaa___ana__________ ________G
A__________nna__an__na___nn__nnn___SSOCIATION_of__ ________G
G__________ana_naa__an___nnn______________________ ________E
N__________ananan___nn___aan_IGGER________________ ________R
A__________nnna____naa____________________________ ________S
A________nnaa_____anan____________________________ ________*
G________anaannana________________________________ ________A
N________ananaannn_AY_____________________________ ________S
A________ana____nn_________IRC-EFNET-#GNAA________ ________S
A_______nn_____na_________________________________ ________O
*_______aaaan_____________________________________ ________C
um, dolor. Nunc nec nisl. Phasellus blandit tempor augue. Donec arcu orci, adipiscing ac, interdum a, tempus nec, enim. Phasellus placerat iaculis orci. Crasa sit amet quam. Sed enim quam, porta quis, aliquet quis, hendrerit ut, sem. Etiam felis tellus, suscipit et, consequat quis, pharetra sit amet, nisl. Aenean arcu massa, lacinia in, dictum eu, pulvinar a

Just imagine,...

[Anonymous Coward]

...a Beowulf cluster of....

Oh wait, it is a cluster. DAMN!!!!!

Re:Just imagine,...

[AchmedHabib]

yeah yeah yeah, but imagine a beowulf cluster of that cluster!

Re:Just imagine,...

[gatesh8r]

*head explodes*

Re:Just imagine,...

[jcr]

Ah, but is it a Beowolf cluster, or a Zilla cluster?

(10 points for anyone who knows what Zilla is.)

-jcr

Re:Just imagine,...

[AchmedHabib]

Someone who is about to get a Cease & Desist order order from Toho Co?

confirmed: I just shat all over myself

[confucio-licious]

I can't believe I just took a shit in my pants at work. What the fuck? I cant stand up and take it to the restroom, because I am wading in a puddle of feces at this very moment. Any change in my seating position will send my poo spilling down my legs and into my socks and shoes. There is no way I can play this off as a fart. It doesn't help at all that I've been spun out for 3 days. I can hardly see straight! How am I possibly going to explain this to my Supervisor when she comes over here to ask me what the fuck I just did in the presence of my employer? She'll make me drop trau in front of everyone again. There it is. The person in the cubicle to the right of me just asked if I could smell "that". I gotta go.

G5 to the power of 1100!!!

[Anonymous Coward]

Worthwhile the wait.

Re:G5 to the power of 1100!!!

[Anonymous Coward]

But will it be as fast as a modern AMD chip?

Re:G5 to the power of 1100!!!

[Anonymous Coward]

Duh, it already stomps them. Now use the IBM compiler and watch dem AMD's burn ;-)))

Agriculture!

[Anonymous Coward]

I suppose this gives new meaning to the term Computer Farm!

Re:Agriculture!

[Anonymous Coward]

Actually, this fits the old definition of Computer Farm....

Re:Agriculture!

[MojoMonkey]

Computer Orchard?

Dear Apple,

[Anonymous Coward]

Dear Apple,
I am a homosexual. I bought an Apple computer because of its well earned reputation for being "the" gay computer. Since I have become an Apple owner, I have been exposed to a whole new world of gay friends. It is really a pleasure to meet and compute with other homos such as myself. I plan on using my new Apple computer as a way to entice and recruit young schoolboys into the homosexual lifestyle; it would be so helpful if you could produce more software which would appeal to young boys. Thanks in advance.

with much gayness,

Father Randy "Pudge" O'Day, S.J.

What?

[thedogcow]

Slashdot posting vaporware news from an unreliable (Thinksecret et al) source?

Get the fuck out of here.

Re:What?

[inkswamp]

Slashdot posting vaporware news from an unreliable (Thinksecret et al) source?

I take it you don't look at Think Secret on a regular basis. It is, easily, the most accurate Mac rumors site out there. In fact, they have posted info on numerous occasions that has caught the attention of Apple's lawyers, and have been forced to pull down and issue their standard disclaimer. Say what you will about other rumors sites (most of them simply feed off each other) but there are some startlingly reliable sources informing Think Secret. Frankly, I don't recall the last time they were wrong about anything they've posted.

Re:What?

[Anonymous Coward]

What about iWalk? And their WWDC coverage was like 50% accurate.

Re:What?

[Anonymous Coward]

"I don't recall the last time they were wrong about anything they've posted"

iWorks.

Re:What?

[Anonymous Coward]

Frankly, I don't recall the last time they were wrong about anything they've posted.

How about this report? I'm still waiting, nearly a fortnight later, after they said "one Apple retail source said the new 15-inch and 17-inch models are expected to arrive later this week".

Re:What?

[ZackSchil]

Heh, the iWalk was all SpyMac. And the WWDC coverage was closer to 98% accurate.

Re:What?

[Trurl's Machine]

How about this report? I'm still waiting, nearly a fortnight later, after they said "one Apple retail source said the new 15-inch and 17-inch models are expected to arrive later this week".

But please observe they accurately noted that this particular information is nothing but an expectation of one retail source. In fact it's quite obvious that no new releases are to be announced just a fortnight before Expo in Paris. Your long-awaited powerbook is most likely to be the Steve's Fabulous "One More Thing" (TM) during the keynote.

Re:What?

[inkswamp]

iWorks.

And you know that's wrong because...? What do your sources say? Did Think Secret announce a release date for iWorks or did they simply report that it was being worked on?

wow man!

[mike77]

dude, imagine a beowulf cluster of... oh.. wait a minute...

Re:wow man!

[karmavore]

Dude, you're getting alot of Dell's!!!!!

Re:wow man!

[Anonymous Coward]

"Dude, you're getting alot of Dell's!!!!! "

No, they're thinking a thousand times different

is that so?

[carpe_noctem]

there's talk of it all over the campus ....and we all know how reliable campus rumors are! C'mon guys, don't forget to say hi to the Olson twins when you see them on campus next year!

Re:is that so?

[Anonymous Coward]

There is a REAL STORY linked asshole.

Re:is that so?

[daeley]

C'mon guys, don't forget to say hi to the Olson twins when you see them on campus next year!

Wow! Imagine a Beowulf cluster of...

Er, ah, forget I said anything.

Re:is that so?

[carpe_noctem]

Wait a sec...you mean the thinksecret story, or the macrumors one? I think "real" is a bit of an overstatement here...

Re:is that so?

[NASAKnight]

It's not just a campus rumor. I attended an informational meeting about it today (which was actually postponed when I got there, so there will be another one soon). Basically, they are recruiting a few computer geeks here at Tech to help set it up and all. Should be fun :P

Stephen

Re:is that so?

[Anonymous Coward]

You mean the Olson twins aren't actually coming?

Re:is that so?

[dbrutus]

So is it going to be a beowulf cluster or an Appleseed cluster?

Re:is that so?

[Anonymous Coward]

they do when i am in between them!

Re:is that so?

[Anonymous Coward]

Quagmire: "I dont wanna come between you........Or do i- Alright!

Re:is that so?

[ReallyQuietGuy]

Er, ah, forget I said anything. it's too late!!!! a beowulf cluster of Olsen Twins!?!?!? OH MY GOD!!!!!

Re:is that so?

[Glock27]

C'mon guys, don't forget to say hi to the Olson twins when you see them on campus next year!

Wow! Imagine a Beowulf cluster of...

Er, ah, forget I said anything.

The *real* reason no ban on cloning will work... ;-)

Re:is that so?

[Wayfare]

Haha - I was there too. Know anything about the new date?

Re:is that so?

[Toe, The]

Who's banning clothing on the Olson twins?

What about latency?

[Thinkit3]

That's the one thing that favors huge amounts of processors in the same box. All this "the internet is one giant distributed computer" doesn't acknowledge this. A box designed to be separate just will not have the latency advantage of a supercomputer designed from the ground up.

Re:What about latency?

[Anonymous Coward]

Who builds 2-8K processors in a box? Have you thought for a second what it might take to power or cool this? Or if it could be wired? The actual engineering of actually doing this are much more challenging than talking crap on Slashdot.

You win the moron of the article award. Congrats.

Re:What about latency?

[andrewl6097]

Actually, the amount of cache coherency checking needed in a large shared memory machine drives the latency right up. Besides, myrinet is ~15 usec latency. That's pretty goot for most things.

Re:What about latency?

[Glonoinha]

-You win the moron of the article award. Congrats.

Now you are one optimistic AC. The day is still young. I am giving 30:1 odds that there are going to be way better morons than Thinkit3 before this thread is archived.

Re:What about latency?

[purdue_thor]

>> A box designed to be separate just will not have the latency advantage of a supercomputer designed from the ground up.
I suggest you look at the list of the top supercomputers in the world. Most are clusters, ie. separate, distinct machines (just a quick glance shows the top 25 all are). It's just too darn hard to make a shared memory computer with 1000's of processors. So the common architecture is to make a cluster of smaller shared memory machines.

Besides, most clusters built utilize special interconnects like Myrinet that offer low latency connections. They're more expensive than ethernet, but it's a supercomputer so you spend it.

>> All this "the internet is one giant distributed computer" doesn't acknowledge this.
On the contrary... people know this very well. That's why we see rendering and SETI processing as distributed. They don't really need to communicate with others often.

Re:What about latency?

[Junta]

There are tradeoffs actually. This isn't like distributed.net or seti@home, this is a controlled network. They have complete control over the network switches, technology, and topology used and can design the network to accomodate tho problems the cluster will be designed to solve.
For example, you could use Myrinet to get 2 Gigabit, super low latency connectivity, or Quadrix, or Infiniband, or just a well laid out Gigabit Ethernet with high end switches.

In multiple processors in a box, the processors have to fight for the resources that box has to offer. NUMA alleviates demand on the memory, but IO operations (when writing to disk or to network) in a multiprocessor box block a good deal as the processor count in a node rises.

The idea with clusters is that inter-node communication in most cases can be kept low. Each system can work on a HUGE chunk of a problem on its own, with its own dedicated hard drive, memory subsystem, and without having too much competition for the network card. A lot of problems are really hard to solve computation wise, but are *very* well suited to distributed computing. A prime example of this is rendering 3D movies. Perhaps oversimplifying things, but for the most part, a central node divides up discrete parts (a segment of video), and each node works without talking to others until done, so the negative impact is minimal. Certain problems (i.e. nuclear explosion simulations where time and spacial chunks interact more with one another) are much more sensitive to latency/throughut. Seti@Home and distributed.net are *extremely* apathetic to throughput/latency issues (not much traffic and very infrequent communication).

Re:What about latency?

[Aapje]

It all depends on what you need the cluster for. Some computations need constant communication, others can go on for hours, days or even weeks without feedback. If you're smart, you use supercomputers for the first kind of tasks and clusters for the second kind.

Universities (and big business) often work together and exchange resources. Virginia Tech gets a large amount of bargaining power by having control over a large amount of processing power. They can easily trade CPU time on their cluster for CPU time on a low-latency supercomputer.

Re:What about latency?

[Kibo]

I wonder if any universities have tried to write a distributed computing app along the lines of seti. Require it to connect to the university network, it grabs itself maybe 50 megs of hd space, and a fraction of all the new computers people bring to campus, in addition to all the computer lab gear belong to their massive number crunching problems. Make another version available to alumni, or even institutions as some form of corporate sponsorship.

Then if it got popular, and they were really clever, they could sell off a part of that computational power they amassed to solve other peoples problems providing for funding for new versions and new supercomputing clusters.

Re:What about latency?

[Anonymous Coward]

Latency is paramount for some tasks, less important for those that *can* make a good distributed project over the Internet of today.

Now, since today's supercomputers are *all* massively parallel constructions, the difference between a commercial design and an off-the-shelf cluster is in the quality and speed of the interconnects. NEC's Earth Simulator, the prime example of 'custom' supercomputer architecture, puts many processor units on *ridiculously* fast 'local' buses, and its racks are all interconnected with still_pretty_insanely_fast (and rather expensive) custom links.

Meanwhile, more 'commercial' designs use various interconnects. IIRC, NEC's 'regular' supercomputers, which formed the design basis for the Earth Simulator architecture, use Fibre Channel 'mesh' networks between racks. The Opteron - sure to be an up-and-coming player in this market - offers HyperTransport, which it looks like Cray will be stretching to its limits on Red Storm; I'm not sure *how* long an HT bus can be, but one gets the impression they'll be stretching it as far as possible, and it's certainly high throughput/low-latency versus the technologies you'd usually find in use for 'networking.'

Anyhow, point is, those designs pack a lot of CPUs together with *very* fast interconnects (equivalent to 16, 32, 64+-way SMP), and have lots and lots of racks of those. (The Opteron/Red Storm approach sounds sexy to me, because I think Hypertransport should let them pack 'lots and lots' of CPUs together versus existing designs. I've yet to read anything about what they're actually doing with it, though.)

Now.. In contrast, an 'off the shelf' cluster is usually going to stick with Ethernet, and will only have 1 to perhaps 4 processors per [node-unit-where-the-CPUs-are-connected-on-a-fast- local-bus], depending how affordable 'cheap' multiprocessor systems are at the time. But *everyone* building supercomputers bumps up against the latency/routing problem; it's just a question of whether it's a problem for, say, 50 Earth Simulator racks (aren't there quite a few more?) vs. 1100 PowerMacs. Experimenting with 'lots of little nodes' has led us to better understand the problem, and learn how to produce tuned topologies that can compete favorably with 'purpose-built' hardware. See: http://aggregate.org/KASY0/

Now, the question *is* one of cost-benefit. Large supercomputers tend to be built with maintenance features and power efficiency in mind. In turn, a totally 'off the shelf' cluster like KASY0 has some advantages because each machine is a cheap, practically disposable 'module' unto itself, and can doubtless be downed off the cluster, pulled out and replaced with another while being easily bench-repaired (since, after all, it's a self-contained PC, rather than a CPU blade or some other random card that would require an expensive test rack to troubleshoot). Meanwhile, if you absolutely demand low-latency, you want one sort of design (Red Storm seems to be acheiving it 'on the cheap,' by combining off-the-shelf - and thus cheap - chips and buses with smart 'custom-design' engineering) while if you can sacrifice some for throughput (jobs with few conditionals), you want another... (like 1100 G5 Macs on a shelf, wired with 'boring' gigabit ethernet, especially if Apple is giving you a bulk discount on the hardware).

So what I'm trying to say is... this is a *combination* of PR stunt and intelligent planning, and there's certainly a lot of 'good science' they could do with the beast - both in number-crunching and 'computer science' a-la cluster topologies. Whether they'll actually *use* it for such, or if it'll be solely a topology toy is anyone's guess.

I think there's some hope that it'll be the "Real Thing," though, since this would explain some of the weird rumors about FC-on-the-mainboard Macs. So they get a Real Monster, made of what will be revealed as "the new G5 Xserves" at the unveiling. The best of COTS *and* fresh d

Re:What about latency?

[BobGibson]

Or how about this: your bandwidth is dependent upon the amount you contribute to the distributed processing.

Hopefully there would be some sort of minimum service level, maybe 64kbps; presumably people dropping tens of thousands expect at least a modicum of return on their investment. People who didn't want to install the client could trudge along at those speeds.

Eventually there would be a market system, whereby people would trade their completed blocks for other commodities, like food vouchers, prints, copies, cash, and sexual favours.

Good luck and godspeed,

Branch

Re:What about latency?

[Anonymous Coward]

Actually, the amount of cache coherency checking needed in a large shared memory machine drives the latency right up.

That might have been true in, say, 1995. It's not true today.

Go look at the cache management design of, say, an Origin 3000. These machines scale to 1,024 processors in the field (NASA has one, and I'm pretty sure there are others), and 4,096 processors in the lab. The latencies are well under the doom-and-gloom estimates you give.

Besides that, the algorithm overhead of PVM/MPT/whatever far outweighs the latency hit in a CC-NUMA system.

Shared memory cache-coherent NUMA is far superior. It's just far more expensive.

Re:What about latency?

[Anonymous Coward]

It's just too darn hard to make a shared memory computer with 1000's of processors.

You can buy a shared memory computer from SGI with 1,024 processors. It's called an Origin 3800/3900. (Depending on which model you get.) It's in the price book. You can just call 'em up and place an order.

They're running configurations with up to 4,096-processors in the lab.

It's not hard to build a thousand-processor computer. It's just expensive.

Re:What about latency?

[Anonymous Coward]

In multiple processors in a box, the processors have to fight for the resources that box has to offer. NUMA alleviates demand on the memory, but IO operations (when writing to disk or to network) in a multiprocessor box block a good deal as the processor count in a node rises.

Seymour Cray once said that a supercomputer is a device for turning compute-bound problems into I/O-bound problems. In other words, if your job is I/O bound, great! That means you're making the best possible use of your compute resources.

In the real world, most jobs require far more compute resources than they do I/O resources. So scaling to a thousand processors or more makes sense, because we can already scale I/O up to gigabytes per second, either disk or network, very easily.

The idea with clusters is that inter-node communication in most cases can be kept low.

Alas, the idea with supercomputers is that inter-node communication cannot be kept low. Consider terabyte-scale data set visualization, for example. There's simply no way to do that job without distributing a copy of the entire data set to every node. That makes is a really bad job for a cluster, but a perfect job for a supercomputer.

Do not be fooled into thinking that clusters are superior to supercomputers, or vice versa. There are tasks that clusters can do cost effectively that supercomputers cannot do as cost effectively. However, there are tasks that supercomputers can do that clusters simply cannot do, no matter the cost. So a one-to-one comparison between the two is inevitably going to be incomplete and misleading.

Re:What about latency?

[stingerman101]

Doesn't it also depend on the interconnect technology they decide on? Since the G5 architecture uses hypertransport for the PCI-X connection and switches the communications to the processor, we are not talking about the same level of latency that older generation PC's had, are we? Isn't it premature to draw conclusions until we better understand how Apple and Virginia Tech plan to architect this new super type computer cluster?

Re:What about latency?

[the quick brown fox]

I found this article on the same site quite interesting.

"By the mid-nineties the new breed of computers made from off-the-shelf commodity chips arrived on the market. Those funded from the US ASCI programme and consisting of several thousand CPUs grabbed the headlines, but because of communication and memory bandwidth limitations, they often deliver very little of their potential peak performance to the user."

Re:What about latency?

[Anonymous Coward]

Doesn't it also depend on the interconnect technology they decide on?

Yes and no. The degree to which a cluster suffers from node-to-node latency depends on the inter-node interconnect used. However, the choice of interconnect cannot remove inter-node latency. When Node X needs to work on data set Y, the cluster controller has to send data set Y to Node X over the interconnect. Data has to be copied into Node X's memory (or onto Node X's disk, if the data set is large). This is not true in a supercomputer. So while you can lower the node-to-node latency, you cannot eliminate it in a cluster.

Isn't it premature to draw conclusions until we better understand how Apple and Virginia Tech plan to architect this new super type computer cluster?

Depends on the conclusion. :) We can conclude, for example, that this cluster will not perform at the same level as a supercomputer with comparable FLOPS, unless the problem can be divided up into data sets that are small enough to fit into a single processor's cache. (Once the data is resident in each node's cache, the difference between cluster and supercomputer disappears, because there is no node-to-node communication at all until the given compute cycle is complete.)

If the data set can be parallelized that easily, then they probably wouldn't use G5's, because the problem wouldn't benefit from vectorization as much as we're all assuming theirs does.

Re:What about latency?

[Anonymous Coward]

although myrinet is great at 250 MB/sec both ways, they will be using Infiniband, which goes up to about 850MB/sec full duplex.

Re:What about latency?

[smallfries]

Dammit! The gambler in me wants to take those odds ... but the slashdot reader is saying I Pity The Fool!

Re:What about latency?

[iantri]

Nice idea, but either this will never happen or their will suddenly be a campus-wide Windows-only policy when it comes time to write the client..

Re:What about latency?

[Anonymous Coward]

You might ask that of Danny Hillis and the rest of the folks that worked at Thinking Machines. They put between 16 and 1024 processors in boxes called Connection Machines and figured out how to power, cool, and wire them. They also figured out how to program them, which turns out to be a whole other kettle of fish, and in so doing produced some of the most powerful supercomputers of their era.

Re:What about latency?

[Kibo]

I was under the impression that there were seti@home clients available for linux and macs.

Re:What about latency?

[mjpaci]

While they were doing this they also forgot about simple things like book keeping and burn rate and are now out of business.

Didn't the CEO or someone very important drive to work in an old fire engine?

Re:What about latency?

[Anonymous Coward]

> Who builds 2-8K processors in a box?

It's been done. The Connection Machine, built in 1988 by Thinking Machines Corporation. System designer and company founder Daniel Hillis employed a hypercube architecture to tie it all together. They achieved 2.5 Teraflops processing power and memory bandwidth of 300Gbits/Sec, all with late-80's technology.

In fact, the CM-2 model incorporated 64,000 processors. And yes, they were all in the same box. So it is eminently do-able.

For verisimilitude, see the very short summary at
David A. Sheppard's site..

And the folks at TMC weren't the only ones to do something like this. Inmos (of England) was producing a hypercube node building block chipset for an architecture they called the Transputer, right around the same timeframe.

> Have you thought for a second what it might take to power or cool this? Or if it could be wired?

You are right to be concerned about thermodynamic considerations, as well as the mechanics and relative performance of interconnection architectures. But you should be advised that there have existed for some time now some rather sophisticated solutions to these problems.

The parent poster's point is anything but moronic, since integration of numerous processors into a single entity reduces time delays and the overhead associated with protocols. For certain kinds of computational problems, this can be significantly important.

> You win the moron of the article award.

So, knowing what you know now, don't you think an apology to the parent poster is in order?

papaTango

Hmmm...

[Undaar]

"Both Apple and the University are playing mum on the issue, but there's talk of it all over the campus."

Must be a pretty boring campus...

Re:Hmmm...

[idiotnot]

Ain't a whole lot going on in Blacksburg, Va. My brother is a physics student there. Wonder if he'll take me to see it if I buy him some beer.

Re:Hmmm...

[holzp]

actually to the average slashdotter it sounds like a fairly happening place.

Re:Hmmm...

[danielobvt]

Must be a pretty boring campus...
Says the person posting a comment on slashdot on the weekend.....
Was that tinkling sound the noise of your glass house falling down? ;)

Re:Hmmm...

[Anonymous Coward]

...says the guy with no life...

Run Photoshop !

[orthopodreloaded]

Are they gonna run Photoshop on that supercomputer ?

Re:Run Photoshop !

[Anonymous Coward]

As it turns out, real men (and this excludes
you) know how to program computers in assembly
to make them run programs other than the
software your mommy and daddy bought you for
your lil' mac computer.

Re:Run Photoshop !

[TheRaven64]

What about X-Code? It supports distributing a compile around a network. I imagine that you could get some pretty sweet compile times with a cluster of that size...

Re:Run Photoshop !

[Anonymous Coward]

Where do you find time to type after jerking
off all those penis'?

Re:Run Photoshop !

[commodoresloat]

Only to show that there are games for the Mac.

1100 G5s still can't...

[Anonymous Coward]

play Doom III.

Re:1100 G5s still can't...

[Anonymous Coward]

I bet 1100 G5's could emulate a PC running Doom III.

Re:1100 G5s still can't...

[Anonymous Coward]

No computer available can really play Doom3. It's not fucking out yet you stupid fucking cunt. Maybe I missed the announcement, that the game was cancelled for the mac. But considering that it was first demoed on the platform, that every single id game has been ported and that you are a fucking moron, I'd say that 1100 G5s can play 1100 copies of Doom perfectly well. Provided they are all equipped with in- and output devices, which is rather unlikely. So I guess you're right, Mr. (or Mrs. or maybe even Ms.) shitface. 1100 can't, but a single one will be able to. In closing I would invite you to remove your head from you stinking rectum and replace it with your dick. If you are female I apologise for calling you cunt, call you wanker instead and invite you to stick a fist up your privates. Thank you very much.

Re:1100 G5s still can't...

[QuantumSpritz]

I'm pretty sure Doom III will run on G5's when it's released. I seem to have read interviews with Carmack that mentioned that fact... ANyhoo.

Dear Father O'Day:

[Anonymous Coward]

Dear Father O'Day:
Thanks for your letter. Being Catholic myself, I know exactly what you're talking about! It has always been our plan here at Apple Computer Inc to revolutionize personal computing with our high-quality and highly gay products.

I'm happy to answer your letter by letting you know that YES we will be releasing an entire hLife ("homo-life") software line. You'll be able to recognize it in stores by the small stylized logo depicting a large cock entering a tight anus with an Apple logo on it. ("Suddenly it all comes together" indeed!).

Anyway, I hope you and other members of our community will join us on our mission, and purchase the exciting new hLife boxed set. Only the boxed set comes with translucent cock rings!

Sincerely,

Harry Rodman
Vice-president
Homosexual Liaison Services
Apple Computer, Inc.

I would like to join.

[Anonymous Coward]

I specialize in fucking gay black dogs. Will this qualify me?

Re:I would like to join.

[Anonymous Coward]

Hey hey momma said the way you move

Gonna make you sweat gonna make you groove

school children speculating ?

[Anonymous Coward]

say it aint so

remember to hand in that homework !

As a VT student...

[Julius X]

but there's talk of it all over campus.

Funny, I haven't heard anything about it prior to today. Guess I'm just out of the loop then...

Re:As a VT student...

[An'Desha Danin]

You're not the only one, I haven't heard squat about it. Might be 'cause I've only been here two weeks, but whatever.

In any case, this is nifty. I have absolutely no idea what they're going to use it for, but if I find it I'm dropping Half-Life 2 onto it.

Re:As a VT student...

[Anonymous Coward]

I work for the University Computing Support department, and I have heared about this over a month and a half ago. Its just that not many people where arund back then, since classes were out of session.

Re:As a VT student...

[Anonymous Coward]

I guess he meant to say there's talk of it all over Apple-users on campus. All 6 of them.

Re:As a VT student...

[zCyl]

Funny, I haven't heard anything about it prior to today. Guess I'm just out of the loop then...

Well, since it's not official yet, it could just be that someone is imagining a Beowulf Cluster. (Wouldn't be the first time.) Hopefully it turns out to be true. The more supercomputers, the better the world.

Re:As a VT student...

[Spleener12]

Actually, they have their own club, and I'm pretty sure you need more than six people to form your own club.

Re:As a VT student...

[Kirby-meister]

As a CS student at VT, I received word of it days ago - Hello all, This email is to serve as invitation and notice of impending Terascale Facility assembly assistance. For those receiving this info for the first time know that Virginia Tech is building a top 10 supercomputer from scratch and we need your assistance. We do have one stipulation to volunteerism and that is you must not be a wage employee of the university. Grad students on GTA/GRA are fine as well as others outside the university that may wish to volunteer. We are expecting to receive machines next week!!! Yikes! In preparation for the assembly process, we need to get volunteers together at the AISB (Andrews Information Systems Building), 1700 Pratt Dr., this weekend. We are planning to have a process orientation session start at 10:00 AM on Saturday, August 30, and last no longer than an hour. We can give you a few more details about the project if you show up and have not been before. :-)

Re:As a VT student...

[CompGeek01]

I got an e-mail about it a few days ago as a CompSci major..

Re:As a VT student...

[Creosote]

Funny, I haven't heard anything about it prior to today. Guess I'm just out of the loop then...

As a University of Virginia staff person, I can tell you that VT's impending purchase of 1100 G5's was announced on our Mac user's group email list back on 28 July. By Apple's regional Higher Education user's group rep, who kiddingly asked when they could expect UVa's purchase order for 1200...

Re:As a VT student...

[FireBreathingDog]

The more supercomputers, the better the world.

But what if they're evil supercomputers? Or what if they're running Microsoft? (-1, Redundant)

Re:As a VT student...

[Anonymous Coward]

Probably because you're a normal student and football starts tomorrow. That's all I've been talking about for the past week, at least. UCF is getting their asses beat!!!!

Re:As a VT student...

[Jaeger-]

Wow as if beating UCF is a great accomplishment. How about picking on someone your own size??
Miami will come to your own house this year & stomp your ass (again as always) don't worry...
But hey keep picking on the little guys if it makes you feel special, and keep posting AC!!

Re:As a VT student...

[dbirchall]

By "running Microsoft," do you mean controlling the corporation, or running Microsoft's code?

If the former, it's probably already happened.

If the latter, I don't think we have anything to fear.

If both... well, don't say I didn't tell you.

Re:As a VT student...

[Anonymous Coward]

Actually, i think it is 3.

Re:As a VT student...

[djward]

As another, I heard of it over a month ago, when they were ordering the machines. A friend of a friend is in charge of hooking the things up. They just started to come in this past week; I'm going to see it next week. I was gonna wait until it was built to subit it to ./, but el gato, el bag...

I'm in Geology, and I knew before most CS/IT type people did. Ha. I guess it really is who you know.

Re:As a VT student...

[mslinux]

Do you know where they are going to put all these machines???

Re:As a VT student...

[djward]

No, actually. I'm kinda curious to see how they arrange them. They're not blades by any means. And 1100!? Ay Caramba! Maybe they'll put them in the power plant as a replacement for the coal-burning furnace. I guess I'll find out next week.

Re:As a VT student...

[confused one]

Ya know, back in '87 when I was on VT Campus and working in the computer support department, there were literally thousands of Apple users. Linux wasn't available. PC's were still primative. The Comp Sci people ran AU/X on Macs. The Physics and Chemistry people ran Macs. Only the Engineering schools ran PC's (DOS machines).

Re:As a VT student...

[confused one]

Evil supercomputers [] running Microsoft

*Shudder* Don't even think about it *Chill*

That's just Hokie

[leviramsey]

And it'll get skunked by 40 teraflops by Duke's supercomputer every year!

Re:That's just Hokie

[TiMac]

And which supercomputer might that be?

I'm at Duke...let's just say I'm "in" on a lot of computing stuff...and I don't know of any supercomputer on campus of any significant magnitude. There's a couple clusters....

Maybe you were just making a joke....I had no idea. :)

Re:That's just Hokie

[Anonymous Coward]

Thank you for playing the stereotypical geek. I believe he was making a basketball reference.

Re:That's just Hokie

[leviramsey]

I was referring to the fact that, as Va Tech is joining the ACC, they'll get beaten twice a year by 40 points by the Blue Devils.

Re:That's just Hokie

[bencvt]

Yeah, in basketball at least... but the football score difference will be about the same. :-)

Re:That's just Hokie

[leviramsey]

True. But then again, maybe another Steve Spurrier will come again and lead Blue Devil football to greatness (well, at least adequacy) again...

Re:That's just Hokie

[Anonymous Coward]

Well, the hokies have a 2nd Vick on their way, so that 2nd Spurrier won't do you any good!

Re:That's just Hokie

[wayward_son]

Only during basketball season. During football season, things may be a bit different.

Problems with my supercomputer.

[Anonymous Coward]

I don't want to start a holy war here, but what is it with you with you G5 zealots? Ive been sitting at my 1100 CPU G5 supercomputer for 20 minutes as it computers a fast fourier transform of an 8Ghz guassian. 20 minutes! At home, on my 60 cpu linux beowulf cluster, the same operation would take 2 minutes if that. Also, while this operation is takiing place, Doom III won't start, and everything else grinds to a halt, even my multithreaded emacs is struggling to keep up as i type this.

My Sun Enterprise 5000 is faster than this machine at times. Super computer addicts, flame me if you want, but I'd rahter hear some inteligent reasons why I should use the G5 supercomputer over cheaper, faster clusters.

Re:Problems with my supercomputer.

[Anonymous Coward]

Doom III won't start

Doom III isn't out yet.

flame me if you want

'kay. Wanker.

but I'd rahter hear some inteligent reasons why I should use the G5 supercomputer over cheaper, faster clusters.

Well, you see, 5 is a bigger number than 4, and intel is stuck at 4, so ours is better. And AMD doesn't even *do* numbers!

Re:Problems with my supercomputer.

[carpe_noctem]

Are you buy any chance the guy emailing out these spams?

Re:Problems with my supercomputer.

[Rosco P. Coltrane]

even my multithreaded emacs is struggling to keep up as i type this.

May I suggest you stop emacs to free resources for your FFT?

Re:Problems with my supercomputer.

[Viadd]

I don't want to start a holy war here, but what is it with you with you G5 zealots? Ive been sitting at my 1100 CPU G5 supercomputer for 20 minutes as it computers a fast fourier transform of an 8Ghz guassian.

And what's with having only 1100 mouse buttons? At the price they're charging, why didn't Apple provide 2200 mouse buttons and 1100 scroll wheels. And why did they use a dead operating system like BSD anyway?

Re:Problems with my supercomputer.

[andrewski]

Your problem is that you wussed out and bought the cheap one.

Of COURSE you need the DP machine! DUH! Are you fucking retarded, or just humorless?

Re:Problems with my supercomputer.

[Anonymous Coward]

oh man. can't we bend the rules a little and mod the parent up to (+10, funny). as an apple user, this has to be one of the bestest posts i've seen in a long time.

Re:Problems with my supercomputer.

[Flamerule]

holy shit, I just died laughing...

Thanks ac, funniest comment I've seen in months.

Re:Problems with my supercomputer.

[Anonymous Coward]

AMD will *do* 64 pretty soon.

Re:Problems with my supercomputer.

[darc]

You idiot! A google search would have solved all of your questions and comments. You should have visited the google usenet groups. alt.supercomputers.help.1100CPUG5, like, duh.

All these newfangled newbies posting dumb questions to ask slashdot. Sheesh. Next thing you know, they'll be asking legal advice.

Re:Problems with my supercomputer.

[Lars T.]

Sure AMD does numbers, they even add a +.

Linpack?

[Anonymous Coward]

Sorry, but I gave up FORTRAN oh, about 25 years ago.

Re:Linpack?

[Anonymous Coward]

Sorry, but I gave up FORTRAN oh, about 25 years ago.

No you didn't. People who are old enough to know Fortran know better than hang around /. and post unfunny comments as AC.

Macs ?

[Rosco P. Coltrane]

Virginia Tech placed the dual-2GHz G5 order shortly after the G5 was announced. Multiple sources said Virginia Tech has ordered 1100 units

Wow, that'll make Apple's quarter for sure :-)

Seriously though, why PowerMacs ? I've always been under the impression that intelloid machines are the cheapest commodity hardware around for equivalent processing power, if not the most exciting. Would anybody know why Powermac G5s are a better choice here?

(Note to computer zealots: it's not a flamebait, it's a genuine question, from someone who is rigorously ignorant of the Mac world. And just in case, the first sentence is a joke, too ...)

Re:Macs ?

[jbolden]

Altivec. Certain types of vector code when compiled to only run on a G4 outperform a pentium even at 3+x the ghz range (i.e. a 800 mhz G4 beating a 3ghz PIV). Assuming similar numbers for the G5 and the increase across the board on all the non vector operations + the fact that the 970 work together so much better....

I can see it making a lot of sense. NASA and lots of bio companies use the G4s this way.

Re:Macs ?

[Mwongozi]

Especially if you're going floating-point based math, the cost-per-crunched-number is actually lower for the dual-G5 tower than it is for a Xeon-based system that costs more anyway.

Re:Macs ?

[WasterDave]

why PowerMacs?

A couple of things make them suitable for clustering:
* There's heaps of processor-processor bandwidth and memory bandwidth.
* On board gigabit ethernet.
* Monster fast execution of properly written vector code.
* Well designed cooling.

Of course, the bang/buck ratio could be an issue for some debate but there's little doubt that in comparison to other commercial unices it's an absolute bargain.

Dave

Re:Macs ?

[damiam]

For certain types of processing (rc5 cracking is one example), Macs completely smoke PCs. For example, distributed.net stats show that a 667Mhz G4 can process more keys/second than a 2.8Ghz P4. Considering how much faster a 2Ghz G5 would be, a 1100-node cluster would be damn powerful if you were doing work that mapped well onto Altivec.

Re:Macs ?

[cperciva]

My guess is that Apple gave them a discount in exchange for the marketing stunt.

Re:Macs ?

[tomstdenis]

This isn't fair though, the P4 sucks. If anything compare the G4 to an Athlon. The athlon has single cycle rotates [unlike the POS P4 processor]. :-)

Also these "G4 vs. P4" comparisons. Do they use the most of both processors or is this an Altivec vs. x86-alu comparison? [or vice versa].

Tom

Re:Macs ?

[Anonymous Coward]

Since Altivec cannnot do double precision there is NO WAY this cluster will end of as number 5 on the top500.

The rules are extremely clear: the ranking is based on high performance parallel LINPACK in DOUBLE PRECISION.

Re:Macs ?

[Llywelyn]

1) As others have mentioned, AltiVec and the dual-FPU on the G5 probably were significant factors in this decision. The Earth Simulator is comprised of processors that are very slow at most tasks, but are designed to scream at vector-optimized code and, honestly, AltiVec makes SSE2 and 3Dnow! look like toys by comparison.

2) You would be hard pressed to configure a dual-opteron or dual-Xeon which could trounce the G5 in terms of speed and cost significantly less. MacOS X server also costs less than any version of windows (pure capital cost here for an 1100 seat license), which may also have factored in.

3) My guess is that they have struck a fairly significant deal with Apple (even so low as Apple provides them at cost, though I doubt its quite that low) in exchange for some degree of publicity when this thing is built.

Re:Macs ?

[Llywelyn]

True, but it is entirely possible that they are building this for more than just to say "we have a computer on the top list and you all sux0r5". They are depending on the dual-floating point units on the G5 for LINPACK and AltiVec for whatever else they put it to.

Without more details its hard to tell

Re:Macs ?

[Anonymous Coward]

Wow, that'll make Apple's quarter for sure :-)

It will also give Apple more bad press due to late delivery if this is true. It's always good with bad. I for one want mine delivered in the near future, thank you. I don't care one bit if Virginia Tech holds up my order. Great logistics Apple! Perhaps a course or two would help.

Re:Macs ?

[be-fan]

That's a *really* bad benchmark. The AltiVec unit has a dedicated vector permute unit which makes RC5 processing results way out of line with the results of most other benchmarks. Its true that the altivec unit is good, but its not *that* good. Besides, both machines have comporable memory bandwidths (if you consider Canterwood P4s) and for many (most?) FP-heavy benchmarks, the code has a "streaming" nature, and is thus almost entirely memory bandwidth limited.

Re:Macs ?

[thebigmacd]

What part of "For certain types of processing" don't you understand? The whole point of the parent post was that certain type. How is a benchmark comparing RC5 crakcing speed a *really* bad benchmark of RC5 cracking speed, or ddn a bad benchmark for ddn performance? You think integer additions is a better benchmark for RC5 cracking speed and ddn performance?

Re:Macs ?

[Anonymous Coward]

"MacOS X server also costs less than any version of windows"

WTF does that have to do with anything? All of the other top500 machines use Unix or Linux, not Windows. Stop trolling.

Re:Macs ?

[be-fan]

"Certain types of processing" (as a phrase) implies that there is a useful subset of the total problem space that the G5 is particularly good at. RC5 cracking is not a useful problem, and thus the G5's performance on this benchmark is wholly unrepresentative even of the types of processing that AltiVec is particularly suited to.

PS> When you use a very common phrase like "certain types of processing" be prepared to deal with all the semantic baggage it brings along.

Re:Macs ?

[damiam]

The fact is that Altivec is really good at some things (such as rc5 cracking), and that G5s are probably the fastest per dollar if those things are what you want to do. How fast they can render CG, serve databases, or play chess is irrelevent if what you want to do is crack rc5.

Re:Macs ?

[thebigmacd]

"When you use a very common phrase like "certain types of processing" be prepared to deal with all the semantic baggage it brings along." Wasn't me who said it :P RC5 cracking is non-arguably a "certain type of processing". There is no semantic baggage.

Re:Macs ?

[11223]

Oh, and I suppose you were asleep during the discussion of the G5's dual, independent double precision floating-point units? And the out-of-order-execution engine that makes them usable? And the memory architecture that will make it scream on large data sets?

The G5's floating point hardware is the most advanced to be found right now, either in standard double-precision or vector double precision.

(FYI: yes, this cluster exists, or will exist. Unfortunately I believe they will be using MPICH which might put a dent into their numbers.)

Re:Macs ?

[andreMA]

I didn't notice in the article (which read, yes) any reference to what OS the nodes of this cluster might run... I think the hardware is really sexy, and I use OS X 10.2 (client) on a daily basis, but i can't imagine that vt.edu is going to build such a cluster and use off-the-shelf software.

Re:Macs ?

[Erich]

Altivec

Doubtful. Altivec can only do single-precision floating point. It's pretty good at it, it can do operations 4 wide, but only single precision. Linpack needs double precision (at least, for the benchmark).

The dual floating point units on the G5 will help, but it's nothing extraordinary. P4s and Athlons both have multiple floating point units. P4's are relatively orthogonal, Athlons less so. However, SSE2 allows for vectorized double precision operations. It is likely that for the linpack benchmark, best-in-class P4 or Athlon architecture-based machines would outperform best-in-class G5 machines.

Altivec is extremely powerful. However it is only useful for applications that don't require their floating point to be double precision. SSE2 is less powerful, but allows for double precision SIMD processing.

Re:Macs ?

[penguin7of9]

Altivec. Certain types of vector code when compiled to only run on a G4 outperform a pentium even at 3+x the ghz range (i.e. a 800 mhz G4 beating a 3ghz PIV).

If that is true, then the G4 and G5 are not very good engineering tradeoffs. Think about it: if the FPUs can only be kept busy by writing very specialized code that 99% of the applications aren't using, then too silicon has been devoted to the FPU relative to the rest of the processor.

And, of course, it doesn't even matter whether some processor outperforms another, what matters is bang-for-the-buck. A single processor G5 machine starts at $2000, more than twice of what you pay for a P4 running at 3GHz.

Re:Macs ?

[JDWTopGuy]

Dude, WTF else can they run? I doubt YDL or any other PPC linux supports the PMac G5 yet. I don't see why they would need anything other than simple OS X client for the slave nodes. Of course they could probably run raw Darwin, but why?

Re:Macs ?

[jbolden]

You can make the same argument for floating point instructions at all 10 years ago (i.e. when you needed an 80387 to do floating point in hardware). Now everyone agrees that the 80486 including the math coprocessor altered the types of code being written for PCs....

I think 5 years from now we'll see tons of code that won't run on the G3 that takes advantage of altivec and in terms of custom code lots exists today.

As for the $2k comparison, I'd like to see where you get an under $1000 machine with: PCI-X, high end video, DVD-R, AGP 8X....

Re:Macs ?

[wrhix]

For truly high speed clusters, the per port cost of the interconnect is a big piece of the puzzle. Thus you still want the maximum per processor performance. If you look at the Top 500 list you will see a number of highly ranked Intel machines. This is a very big recent development but they are all Xeon and Itanium systems, not budget Pentium 4s. So the G5 is price competitive, especially for floating point apps where PowerPC has always shined.

It would be a big coup if Apple can pull off even a top 10 finish, and it's not out of the question. If you'll look at the second decade of the Top 500 list you'll see several ~1000 processor pSeries 690 Turbo clusters. These are all 1.3GHz Power4 machines, the sister chip of the PPC970. Double the number of processors, each with roughly 1.5x clockspeed, and you're in the ballpark of a top 5 finish. The recently announced IBM FORTRAN compiler will help a lot with this. The open question will be whether the interconnect, which is rumored to be Infiniband, can keep up. If it works, this will sure sell a lot of G5 Xserves!

Re:Macs ?

[Snocone]

Seriously though, why PowerMacs ? I've always been under the impression that intelloid machines are the cheapest commodity hardware around for equivalent processing power, if not the most exciting. Would anybody know why Powermac G5s are a better choice here?

Well, three physical big ones that spring to mind off the top of my head:

a) Space limitations -- better MIPS/ft^3 with PPC
b) Power supply restrictions -- better MIPS/watt with PPC
c) Environmental considerations (PPCs take high temperatures a whole lot better, as a general rule, and produce nowhere near as much heat for a given MIPS capacity to boot)

Of course, depending on the software you're running, as other people have already pointed out Altivec code may be capable of wiping the floor with any other architecture on the planet by pretty much any metric you propose.

If I thought a little more I could come up with some more esoteric reasons I'm sure, but those would cover all the major reasons that $/MIPS is not the only metric worthy of considerations I imagine.

Re:Macs ?

[Anonymous Coward]

I think that five years from now Apple will still have 5-10% of the market and 'AltiVec' will be another retired buzzword. I mean, the DEC Alpha was a hella processor, but only specialized code was written to take advantage of it, and now it's dead.

Re:Macs ?

[The Infamous Grimace]

...or any other PPC linux supports the PMac G5 yet...

IBM makes the G5 (PPC 970), and they use Linux. Betcha there's at least one version of the penguin that'll run on 'em.

(tig)
"We do not inherit the land from our ancestors"
"We borrow it from our children"

Re:Macs ?

[evilviper]

I've always been under the impression that intelloid machines are the cheapest commodity hardware around for equivalent processing power

Yes, the machines are the cheapest. They are also the most power-hungry, and output the most heat. More than double compared to an equivalent-speed G5 (according to the specs I've read).

It's not all that unusual to pay more up-front for systems that will save money over time. Frankly, it's unfortunate that more people don't think about that.

Besides, a while back on /., IBM said they would eventually be releasing a quad-processor PowerPC (equivalent to G5) system rather inexpensively (about 3.5K IIRC). That would put it surprisingly near to Intel systems in price/performance, so the money saved on power and cooling would be recouped much quicker.

Maybe they have a special deal with Apple, to get this stuff cheap? That would tip things in favor of PowerPC systems as well.

Re:Macs ?

[crayz]

I don't see why they would need anything other than simple OS X client for the slave nodes.

Uhh, maybe because OS X has way more overhead than you need for a computer you're just using for its CPU

Re:Macs ?

[Anonymous Coward]

Did Steve Jobs send out a fax saying you Apple advocates are supposed to say Xeon as often as possible.

We know that the only way to price compare the Macintosh and come out ahead is to compare it to the most expensive hardware it competes against. I remember back in the day when Apples were always lined up against Compaq boxes. That was back when Compaq boxes were the most expensive x86 hardware on the market, after they'd priced themselves up over IBM's boxes.

Re:Macs ?

[anthonyrcalgary]

>> "Of course, the bang/buck ratio could be an issue for some debate but there's little doubt that in comparison to other commercial unices it's an absolute bargain."

Apple may have given them a price break for the PR and tax writeoff.

Also, G5's are expensive in comparison with Athlon's and Pentiums, but against Xeons and Opterons it's different. And a lot of these clusters are being built with the higher end parts to reduce the TCO.

With IBM's XLC compiler, I don't see the G5 having much trouble keeping up if they stick to floats. Opteron's memory architechture gives it the edge in a lot of areas, but as the saying goes "The G5 is a floating point monster.".

Re:Macs ?

[Anonymous Coward]

I'm wondering if the boxes all have Harmon-Kardon speakers as was a big marketing-hype point in the days of the early iMac.

Apple marketing loves stunts and tie-ins.

Re:Macs ?

[penguin7of9]

You can make the same argument for floating point instructions at all 10 years ago

We already knew ten years ago that built-in floating point support was good for desktop machines because pretty much all architectures other than x86 had them. x86 just finally caught up with the rest of the world.

Vector processors, on the other hand, have been around for decades and, if anything, they have become less popular. That's because most CPU vendors have figured out how to let people achieve the same kind of performance with regular instruction sets.

I think 5 years from now we'll see tons of code that won't run on the G3 that takes advantage of altivec and in terms of custom code lots exists today.

I think 5 years from now, IBM will either have figured out how to get full speed out of their FPUs without special AltiVec instructions, or the whole PPC architecture will end up in the dustbin of history.

As for the $2k comparison, I'd like to see where you get an under $1000 machine with: PCI-X, high end video, DVD-R, AGP 8X....

Who cares? We are talking about compute clusters; the machines don't need DVD-R or high-end video. The fact that the G5s only come in gold-plated, fashionable Apple desktop boxes is another problem with them for scientific computing.

Re:Macs ?

[IM6100]

Why would they run MORE than Darwin raw for the kind of tasks a cluster would be used for?

Re:Macs ?

[anthonyrcalgary]

I highly doubt Apple charged them a cent for the OS license. Apple can get a tax writeoff for the sticker price of 1100 copies of OS X server unlimited client, but they can print the CD's for next to nothing. Donating them makes a LOT of sense.

Besides, if VT had to pay for the OS they would just take the hardware and use Linux.

Re:Macs ?

[ruprechtjones]

"Without more details its hard to tell"

Bingo. Just as with all the benchmarks coming out about the G5 vs. the Intel/AMD world, nothing matters until this is implimented and benched in the real world. Time will tell.

Re:Macs ?

[Anonymous Coward]

a) Space limitations -- better MIPS/ft^3 with PPC

You are not seriously suggesting that DESKTOP dual G5 machines take less space than 2U rackmounted dual Xeons?

I've been involved in several cluster constructions, and there is one and only one thing that determines deals like this: the price. VT almost certainly negotiated with multiple vendors (like we've done :-), Apple saw their chance for a showcase, and offered a price the other vendors weren't willing to match.

Nobody in academia pays anything close to list prices for big clusters. Although we all have preferences when it comes to vendors and architectures, it is very hard to motivate going with vendor X for $2M if vendor Y offers the same performance for $1M.

Re:Macs ?

[burns210]

now, if only they would use this supercomputer to run the Distributed.net client!

Re:Macs ?

[Anonymous Coward]

but there's little doubt that in comparison to other commercial unices it's an absolute bargain.

Unix is not a processor.

Re:Macs ?

[Lars T.]

You have no clue what SIMD is, do you?

As for the setup of the machines - even going through the Apple Store you can change the DVD-R for a Combo. I guess with a 1100 machine order, Apple lets you choose more (or rather less) than just that. Which brings us to the desktop boxes - many of the PC clusters in the Top 500 Supercomputer Sites list are build with plain-jane desktop PCs, why can't one build one with "gold-plated, fashionable Apple desktop boxes"? Because you say so?

Re:Macs ?

[Anonymous Coward]

You have no clue what SIMD is, do you?

Sure, I do. You have no clue what engineering tradeoffs are, do you? If a G4 only achieves P4 performance on a small fraction of programs and costs the same, it's a bad engineering tradeoff.

why can't one build one with "gold-plated, fashionable Apple desktop boxes"? Because you say so?

You can do whatever you like. The question is whether it's cost effective.

Re:Macs ?

[Lars T.]

Apple's list of (third party) clustering tools.

Re:Macs ?

[Lars T.]

Heat. CPUs create quite a lot, some more some less (like the G5). A few years ago a German university (Chemnitz IIRC) build a large Linux cluster. For cost and speed reasons they wanted to go with Athlons, but decided to use P3s instead, mostly because they created less heat. It is one thing to pack a couple of hundred computers into a building, it's an other to also buy a new air-conditioning for that building.

Re:Macs ?

[Lars T.]

So no programs that are fast on the P4 use SSE2? Is that your final answer?

Re:Macs ?

[BiggerIsBetter]

What's MPICH?

Re:Macs ?

[Magnus Reftel]

Would anybody know why Powermac G5s are a better choice here?

You've already got quite a lot of possible answers to that question, but here's another one: The G5 is a stripped-down power4, and for some reason, the power series is rather popular with the super computing crowd (about 50% of the systems in the upper parts of the top500 list are based on power, 20% in the entire list).

Perhaps they see the G5 as a way to get much of the benefits of a real power4, but at a more sensible price?

Re:Macs ?

[Anonymous Coward]

So no programs that are fast on the P4 use SSE2?

Your question seems a bit confused. I assume what you mean is "does the use of SSE2 result in any improved performance for at least some P4 programs"?

I assume it does, otherwise Intel wouldn't have included SSE2. But the improvements using SSE2 are probably quite modest, and nowhere near the several-fold increases people claim for AltiVec, transforming a pokey 1GHz G4 (roughly the equivalent of a 1GHz P3) into something that supposedly beats a 3GHz P4. Such huge performance increases would indicate that there is something rather unbalanced about the G4 processor design.

And even if AltiVec could achieve such miracles, it would be useless for most compute cluster applications because computer cluster applications are usually scientific applications using double precision floating point and AltiVec is single precision only. SSE2 actually includes double precision instructions.

Re:Macs ?

[Ninja Programmer]

AltiVec would be illegal to use for the Linpack test. AMD/Kentucy tried this with 3DNow! in the KLAT2 machine several years ago but were rejected because the computation was not in 64 bits.

AltiVec cannot perform 64bit floating point computations, ergo, AltiVec will not be a factor.

That said, Opertons and Pentiums both have SSE-2 which *does* support 64bit floating point. Thus the question still remains -- why use G5's, when x86s are both cheaper *AND* faster?

Re:Macs ?

[Ninja Programmer]

why PowerMacs?

A couple of things make them suitable for clustering:
* There's heaps of processor-processor bandwidth and memory bandwidth.

They use HyperTransport which was invented by AMD and used in Opertons.

* On board gigabit ethernet.

Huh? Such things exist in the x86 world as well.

* Monster fast execution of properly written vector code.

AltiVec cannot be used; it can only perform 32 bit floating point calculations which is not legal for the Linpack benchmark used at the top500 site. They have to stick to ordinary FP usage, where they will not have any advantage over Pentiums or Opterons.

* Well designed cooling.

Well designed in terms of what? They are *CLUSTERING* many machines. What matters is the *aggregated* cooling solution, not the cooling of any one single node. The Apple G5 has far more fans than your typical x86 box, which in the end just means more parts which can possibly fail, which becomes a noticable factor in clustering solutions.

Of course, the bang/buck ratio could be an issue for some debate but there's little doubt that in comparison to other commercial unices it's an absolute bargain.

Oh yeah -- they can't strike out the cost of buying a copy of MacOS for each machine, can they? At least with x86's they only need to buy or download one copy of Linux.

Re:Macs ?

[Ninja Programmer]

... honestly, AltiVec makes SSE2 and 3Dnow! look like toys by comparison.

That seems a little hard to believe considering that AltiVec can only perform 32 bit floating point (which is not valid for the LinPack test used in the top500 list) and SSE-2 can.

Re:Macs ?

[Lars T.]

Your answer seems a bit confused. Face it, without SSE2 the P4 would stink if it weren't for the fast clock. Heck, you wouldn't even have good FP performance, or why else would Intel tell everyone to use SSE for it instead of the old FP unit?

The fact that AltiVec actually scales pretty well for many tasks doesn't prove that the G4 is "unbalanced", it shows that SSE2 isn't particularly good. BTW, only double precission? Tsk, you don't seem to need exact results.

Re:Macs ?

[Ninja Programmer]

The G5's floating point hardware is the most advanced to be found right now, either in standard double-precision or vector double precision.

AltiVec does not support vector double precision.

Re:Macs ?

[Snocone]

You are not seriously suggesting that DESKTOP dual G5 machines take less space than 2U rackmounted dual Xeons?

That's not the "cheapest commodity hardware" OP was referring to anymore then, is it? The appropriate comparison then would be to a G5 XServe. And chances are the statement would still hold, I'd say. Although you could certainly pick a specific task to prove whatever point you wanted to make, I'm sure.

Re:Macs ?

[jellomizer]

There is also a good chance that they got a good discount for those 1100 Power Macs. And the fact that they wanted to make the fastest computer on the earth I am sure the PR people at Apple decided to give them some more dicount so they can say that they have one of the fastest computers on earth. Most of the other comanies the Dells and Gateways etc. probably wouldn't give such a discount for the name because Intel would just get the credit for the acheavment and not pc manafucture.

Re:Macs ?

[11223]

Sorry, that was a think-o. If you read it it's obvious I knew that and meant to say single there.

Re:Macs ?

[jbolden]

If a G4 only achieves P4 performance on a small fraction of programs and costs the same, it's a bad engineering tradeoff.

Or its an older design. The G4 came out before the P4's. At the 500mhz range the G4 is comparable 800mhz PIV (is such a thing existed) for general applications. The G4s had a bus problem and failed to keep up with Moore's law. The result being that only in very limited areas were they better chips. The G5s OTOH are much more modern.

Re:Macs ?

[mslinux]

Would anybody know why Powermac G5s are a better choice here?

They aren't a better choice. This is nothing but a *huge* waste of money by the CS Dept. Head who is a *classic* Mac fanatic that has been influenced by the Steve Jobs reality distortion field. The drinking of the poison koolaid is next on his list of things to do right after making x86 a forbidden platform in the dept.

Re:Macs ?

[Anonymous Coward]

You are right, Altivec can't be used since it is only single precision. You assumption of x86s being faster is incorrect however. In the case of the Opteron, the main FPU is a dual double precision unit incapable of fma. The first law of compute processor FPU design is that you need to support single issue fused multiply adds, because you get 2 flops/cycle throughput, which is twice that of the Opteron. For the same performace, you need twice as many Opterons as PPC970s, which makes them non-price competitive. SSE2 doesn't help much. Try coding in it and see the impact. Also take a look at BLAS, the basic math library routines for the Opteron, and you'll see why. To top it all, the 2GHz PPC970 is clocked higher than the fastest Opteron out there. In the case of 32 bit x86 platforms, they have a different problem. It is memory bandwidth. To the first approximation, you cannot compute faster than the throughput of the memory subsystem and the Xeons and P4s have less than half the memory bandwidth of a G5. Interestingly, the Opteron is faster than even the G5 in memory bandwidth, but the see the issue above. On the positive side the 32 bit x86s are clocked very high, which is what makes them even competitive in this race. The processor to look at is the Itanium2, which has similar FPU capabilities as the PPC970. The problem is price (>10 times more expensive) and the clock rate (1.5GHz vs 2). Also, compilers on Itanium2 are a problem. It is not easy creating a optimizing VLIW compiler. Thoughts?

Re:Macs ?

[switcha]

Seriously though, why PowerMacs ?

OK, if a G5 can blast you through and exterior wall and two interior walls of your house, just imagine the sheer wall-blasting power of this many G5's.
Intelliod? That's just computing. G5's? I think we've found the WMD.

Re:Macs ?

[tmattox]

I am one of the designers of KLAT2 and KASY0, and the guy who ran the Linpack benchmarks on both. Over 3 years ago when we submitted our results for KLAT2 to the top500 list, there was no public indication that 64-bit floating point was required. It took them awhile, but the top500 website now has a FAQ that indicates "full precision" is required, and they interpret that as 64-bit for most machines. FYI, 32-bit FLOPs are useful in many situations, and machines had been on the top500 list that had used 32-bit FLOPs. You might take a look at our KASY0 FAQ on GFLOPS. As a means to rank the top500, I think it is quite legitimate to require 64-bit FLOPS, but that doesn't make it "illegal" to use 32-bit Linpack FLOPS for other comparisons.

As for the G5, it won't need AltiVec to get good Linpack numbers due to its fused multiply-add capability in its dual floating point pipes. That's 4 FLOPs per clock peak! I hope VT was able to get Apple to leave out, and not charge for, the components not needed in a cluster node. The PCI-X slots in the G5 should allow VT to better use a high-speed cluster network technology. Commodity x86 boxes tend to only have 32-bit 33MHz PCI, limiting the usable link bandwidth between nodes to under a gigabit per second. For 64-bit Linpack GFLOPS per dollar, a cluster of G5's could be competative. I look forward to seeing their results, and any similar work using the upcoming Athlon 64.

Re:Macs ?

[WasterDave]

The question was, "Why PowerMacs". Not "why not Opterons". There's nothing wrong with Opterons, AFAIK, and certainly in any bang/buck discussion the Opteron is going to float to the top.

The 32bit fp on AltiVec thing is interesting though. I have a suspicion there is going to be some egg on face over that.

Dave

Re:Macs ?

[afantee]

>> Would anybody know why Powermac G5s are a better choice here?

To start with, Intel doesn't offer a 64-bit CPU remotely comparable to the G5 in price. The Itanium 2 costs $3000 per chip, which is more than the price of the dual 2 GHz Power Mac that comes with GPU, RAM, SATA HD, Firewire, USB, Gigabit Ethernet, ...

Furthermore, the dual 2 GHz Power Mac beats the dual 3 GHz Xeon Dell in both 32-bit performance and price ($3000 vs $4000), and of course the G5 can do 64-bit computing that the Xeon can't.

In short, the G5 is cheaper and more powerful than x86, and also future-proof.

Re:Macs ?

[Scudsucker]

They use HyperTransport which was invented by AMD and used in Opertons.

So what? Did he say it was exclusive to PowerMacs? And Apple is a member of the HyperTransport consortium.

* On board gigabit ethernet.
Huh? Such things exist in the x86 world as well.

But they usually aren't standard on the motherboard.

AltiVec cannot be used; it can only perform 32 bit floating point calculations which is not legal for the Linpack benchmark used at the top500 site.

Wonderful. Once you've gotten your "top 500" rating you can turn it back on and start spanking other clusters. Next?

The Apple G5 has far more fans than your typical x86 box

Which was done to reduce the amount of noise, not because the 970 puts out enormous amounts of heat.

Oh yeah -- they can't strike out the cost of buying a copy of MacOS for each machine, can they?

Hmm, lets think about this for a second. They're buying over a thousand brand new machines and giving Apple some great PR. Do you really think that Apple wouldn't come down on the price of the included operating system?

Any more jerky questions?

Re:Macs ?

[Anonymous Coward]

Hi Tim, I fully agree with you, 64 bit fma kinda helps this processor. I also went out and did some costing, right now the dual G5s are the cheapest platform/flop. Sounds rather weird, since Apple is not exactly known for a cheap platform. I agree, it should be really interesting to see what they really get on LINPACK. Till then, it is just informed speculation.

Re:Macs ?

[Anonymous Coward]

Hi Tim,

You have my respect for admitting 64-bit is a reasonable standard. As you say, single precision can be useful, but the problem in my experience is that it requires you to be careful with the order of summation and taking the difference of two big numbers, and it's not always worth the time to rewrite/check software.

For an average user that doesn't sound to difficult, but it is when you have tens of thousands of lines of code that need to be revalidated in you switch to single precision... Programmer time is money, especially when the programming has to be done by scientists.

Incdentally, IMHO Linpack isn't a very good general purpose benchmark either, since very few applications will be able to pack multiply and add instructions in every cycle.

In theory, a single 2GHz G5 with dual floating-point units and FMA instructions is capable of 8 GFlops in double precision. Using gcc (which doesn't generate SSE2 by default), a single 3GHz Xeon is only capable of 3GFlops in double precision, but still score essentially the same in specFP.

Re:Macs ?

[Anonymous Coward]

Your answer seems a bit confused. Face it, without SSE2 the P4 would stink if it weren't for the fast clock. Heck, you wouldn't even have good FP performance, or why else would Intel tell everyone to use SSE for it instead of the old FP unit?

Speaking as somebody who has written a lot of both Altivec and SSE code, it is a bit more complex than that. Even the intel compiler doesn't generate SIMD code in most circumstances. However, the SSE/SSE2 instruction sets include instructions that are essentially normal floating-point operations operating only on the first element of the XMMn registers.

The reason why Intel tells you to use SSE instead of x87 is that they have focused on optimizing the SSE unit and only provide the x87 registers for compatibility. This way all code will be fast (even if it is "normal" FP code and only uses the first element), and in the very few cases where the compiler or programmer can optimize and use all elements it will be very fast, without having to move stuff between the normal FP registers and the XMM ones.

Go ahead and compile with -S to see the assembly output - it's quite interesting.

The fact that AltiVec actually scales pretty well for many tasks doesn't prove that the G4 is "unbalanced", it shows that SSE2 isn't particularly good.

As a theoretical pissing contest about the cleanest vector implementation I agree that Altivec is much better. The problem is that most problems aren't perfectly vectorizable, and as soon as you need to access unevenly spaced (not only unaligned) data Altivec sucks.

SSE, on the other hand, is an ugly hack just as a lot of other things from Intel. But just as most of the other ugly things from Intel the performance often wins over a theoretically nicer architecture in practice. For general algorithms SSE can often be better than Altivec. Intel isn't stupid - they've got some of the best engineers in the world.

It isn't a religion: OS X and Apple aren't bad just because SSE can be pretty good in practice. However, I do agree with the statement that the G4 is a bit "unbalanced" in that it is very fast for a small set of programs that have been manually tuned with Altivec, but it is very slow (especially on double precision FP) on the huge majority of compiled code.

In 20 years I think we're all going to be using the type of explicitly parallel instructions as present in ia64. It is horribly difficult to program, but it allows you to get much closer to the theoretical peak, and you can increase performance with more integer/fp units instead of only higher frequency. (Normal compilers can't really schedule more than 2 fp units).

Again - don't underestimate Intel. They've got WAY too much money, patents, & prestige invested in ia64 to let it fail. With Madison they have just shown us they can produce the fastest CPU in the world - all that remains now is getting the price down, and that's economy of scales (read: a matter of time).

Re:Macs ?

[Anonymous Coward]

For certain types of processing (rc5 cracking is one example), Macs completely smoke PCs.

Actually, RC5 is almost the only, and a very particular example. The Altivec instruction set includes a special rotate instruction that is essentially only useful for RC5 cracking. Obviously, this means it's going to perform very well in this case, but the reason nobody else has implemented it is that it is useless for general purpose computing.

It doesn't make the RC5 result less impressive, but it is by no means representative of Altivec performance improvements.

Re:Macs ?

[Anonymous Coward]

Most of the other comanies the Dells and Gateways etc. probably wouldn't give such a discount for the name because Intel would just get the credit for the acheavment and not pc manafucture.

Actually they do - and Intel discounts the CPUs for them :-) I can't give you any details (NDA), but big academic installations often pay about 25% of list price. A cluster is much more about administration and total cost of ownership. One of the most common solutions is to use rackmounted dual Xeons with remote access cards (so you can even access a completely dead node and "push the power-on button" remotely, and get reports on hardware problems). With 1000 nodes it is really useful when a failing node has a big red light on the front, and an LCD saying "FAN1 RPM LOW". Slide the node out, open the lid, replace the fan module with a new one and have it up and running 5 minutes later.

It is never a matter of first selecting G5, P4 or anything else. You put up a specification, collect benchmark statistics on the most important software you plan to run, and then negotiate on the price and other things (like the vendor being able to use it in advertising).

My bet in this case is that Apple wanted to use the momentum when the chip was new (and competitive) and for the first time get a computer into TOP500 to prove their platform. If VT payed more than $1000 per box they made a lousy deal.

Re:Macs ?

[jcr]

Altivec can only do single-precision floating point.

Not quite correct. Apple has extended-precision libraries available for Altivec.

-jcr

Re:Macs ?

[Jungle guy]

These G5s processors have 64 bit processors and , what is more important to supercompters, 1 GHz frontside bus (one for each processor). However, it seems like an expensive choice for a supercomputer, as you have a lot of expensive accessories integrated on the board or bundled on the computer that are not used on a cluster - a powerfull Radeon graphics card and firewire 800 and 400 port. Apple also is not well know in the supercomputing field, and I wouldn't be surprised if this university picks a Linux kernel with a Mosix patch. But, when IBM starts to sell computers with the G5 processor and puts their team to develop clustering software for it, the Itanium and the Opteron will have a very tough opponent.

Re:Macs ?

[cosmo7]

MPICH is a freely available, portable implementation of MPI, the standard for message-passing libraries.

Re:Macs ?

[sean23007]

Bear in mind that a lot of the Top 500 supercomputers are based on IBM Power series processors, and that the G5 (PPC 970) is a cheaper version of the Power4. It is quite reasonable for them to buy Macs for this cluster, since they are virtually the same as Power 4 processors from IBM for less, why would they want Intels when they could get IBM Power processors for a bargain.

Re:Macs ?

[jcr]

AltiVec cannot perform 64bit floating point computations

Not correct. Altivec can indeed perform 64-bit floating point operations, it just can't do them in a single pass.

-jcr

Re:Macs ?

[djoiner]

What do you have against MPICH?

If they are running using any of the standard OSs available for PPC, and the software currently available on those OSs, they're stuck with MPICH or LAM or something else that doesn't optimize collective communications.

Other than MPICH or LAM, what would you recommend?

Re:Macs ?

[Erich]

Not quite correct. Apple has extended-precision libraries available for Altivec.

Go look up the altivec instruction set and tell me which instructions work on packed double-precision values.

You can do double precision on PPC, but you don't get anything from the vector unit. Only the FPUs.

Re:Macs ?

[jcr]

Go look up the altivec instruction set and tell me which instructions work on packed double-precision values.

As I said, Apple's published extended-precision libraries that use Altivec. You can indeed use altivec for double precision operations, you just have to use multiple passes.

You can do double precision on PPC, but you don't get anything from the vector unit. Only the FPUs.

You are mistaken.

-jcr

Re:Macs ?

[Erich]

As I said, Apple's published extended-precision libraries that use Altivec. You can indeed use altivec for double precision operations, you just have to use multiple passes.

So, how exactly is this implemented? I'm very curious how they can do 64-bit floating point efficiently without 64-bit floating point instructions.

Please give me a link or something to how it's done, "multiple passes" makes no sense (to me).

It should be enough

[DarthVeda]

It's a G5 right? Then it should be enough to run Wolfenstein. Oh did they release that for MAC? I can't remember.

Re:It should be enough

[IM6100]

They released Wolfenstein 3D. Long after we'd all played it out on our '286s and '386's, of course. The Mac version of Wolfenstein 3D has a better music track.

Re:It should be enough

[Anonymous Coward]

Actualy, the 3DO version of Wolfenstein 3D was the best offering I have ever seen: superior music track and every episode of Wolfenstein 3D all roled into one CDROM. Why didn't ID Software do things like that? Still, there is not a multiplayer Wolfenstein 3D :(

Re:It should be enough

[Anonymous Coward]

when you had wolfenstein, we had pathways from bungie. when you had doom, we had marathon, also from bungie. marathon.

Do they have a need for it?

[John Seminal]

I might be wrong, but I think VT is a public school. Why do they need the 5th most powerful cluster in the world? How about lowering tuition? Does the school really not need anything else more than a cluster? In 5 years, will it still be worth the money they are paying? Is there an immediate need to the cluster?

Re:Do they have a need for it?

[Anonymous Coward]

I would assume that if this is true the money came from some sort of donation (specifically for this), a grant (again, specifically for this), or some other means outside of money that is allotted for student funding.

Re:Do they have a need for it?

[harks]

Schools have many different accounts set up to fund many different things. This is due to how donors donate money and specify that they want it to go toward a certian project or department. One department, say the CS department might have recieved donations from CS alumni. Also, having large projects like this can generate lots of revenue through grants.

Re:Do they have a need for it?

[absoluthokie]

Exactly, Virginia tech has a goal to become a top 30 research university. Having known about the plan for some time, this makes perfect sense. The departments who are building the cluster, have gotten very large grants and donations from our great alumni to build this, and become a better university for it. This construction can be compared to the stadium expansions. The stadium expansion is paid out of a different set of funds, as is research. Academic fund is hurting because alumni rarely give money for academic reasons, but more for football or research.

Re:Do they have a need for it?

[Kibo]

The grant money that flows into a public research and occasionally teaching institution can be stagering, and absolutely dwarf the money students pay in tuition (sometimes by a factor of 10!). A better question might be, why don't the gradstudents donating their labor, possibly to patents that will be controlled by the university, recieve more consideration, and fair labor law protections.

But I would bet this will be not too dissimilar in use from the HP Itanium2 referenced earlier on slashdot. I would bet one of the paramount concerns this cluster would look at is the effect of farm runoff, and probably climatology too among other things.

Re:Do they have a need for it?

[jabberjaw]

Perhaps for modeling various atmospheric situations, helping out in the Physics department if the campus happens to be doing any research in... well damn near any area of physics would benefit from having a supercomputer handy. The applications in the CS department alone are staggaring. If you were a student of VT, wouldn't you like access to this kind of hardware. I sure would...

Re:Do they have a need for it?

[Kalak]

President Steiger's stated intention his first year in office was to be in the top 25(?) research instutions. The decision to spend the money on this is probaby connected to that initiative. I'm not sure where the fuding came from, but I doubt they could do this without a lot of approval from the top.

Re:Do they have a need for it?

[KiahZero]

Obviously you haven't looked at VT recently. Tuition and fees is only $7,500, out of state. I can only wish that my tuition were that low. Hell, for in-state students, the room and board is the same price as tuition (around $2,000). But of course, you're modded up insightful, because you pulled a random idea out of your ass and presented it as fact.

Re:Do they have a need for it?

[John Seminal]

You are an idiot. I guess since the tuition has not exploded at VT, then it is okay to spend at will. I guess the question of need is not in your vocabulary. Just buy it, even if it is not needed. Jerk.

Re:Do they have a need for it?

[John Seminal]

And you think $14,000 a year for out of state tuition is cheap. It still comes to over $20,000 a year for study. Stop posting numbers for semesters when you obviously imply year.

Re:Do they have a need for it?

[Anonymous Coward]

Yea, that OUT OF STATE SHIT IS BULLSHIT I'M PAYIN LIKE 20 GRAND a semester. It's like 4 grand just for room and board.

Re:Do they have a need for it?

[John Seminal]

Good points. But my question is do they need such a large cluster to accomplish those objectives? Could those objectives be accomplished with a smaller cluster (and having money left over)? And if the cluster was smaller, then the money could be used on other university improvements- perhaps a wireless campus or better student lounges or better food service? I guess the reason I brought it up is because when I was in college, they were charging us $1.20 for a can of pepsi from a vending machine, and that made wonder if money was being wasted.

Re:Do they have a need for it?

[Kalak]

Start looking at prices for other states, and you'll realize that Virginia Tech is cheap compared to other places. If you went here and talked to anyone from the North East, they'll probably tell you what costs are like up there, and that it's cheaper out of state at VT than up north. (Quick check of PennSt. just got me $6100/semester IN state. Check some others up north and you'll probably find it gets higher quickly.)

Also, it's common to give college costs by semester, as saying $20k/year isn't right either as it doesn't include summer classes, which many take (and not just because they fail, but to finish early).

I'm still paying for my college, and it helps to know what you're paying for. Or you can just whine about it on /.

Re:Do they have a need for it?

[steve_bryan]

So how does this anti-intellectual tripe qualify as insightful? Any yahoo can point at and complain about just about any non-trivial project at a research university whether it is public or private. If they were building it just to attain a certain ranking without any research proposals or plans it wouldn't be hard to find fault. Does anyone that could possibly be the case here? I think this sort of empty headed bushwhacking is a cheap shot and contemptible.

Is there something particularly about building any clusters today that is ill advised? Anything specifically about a cluster built with these parts? Why do any science that involves a large expense when the money could be applied to "lowering tuition"? Maybe because an important part of the mission of some universities is to advance the state of knowledge by performing research that would not be done by other segments of society.

Re:Do they have a need for it?

[JohnsonWax]

Well, people also need to realize that universities slut out their resources from time to time. My school has a fabrication facility that is almost exclusively supported by industry. Basically, if they need a $1M piece of equipment (common) in the facility, industry buys it for us and can come and use it. Of course, we use it when they're not, and we rent out time on it to other companies when we're not using it.

The upside for industry is that they don't need to maintain a 100/1,000/10,000 class facility 24/7, and they get to rent out other people's $1M pieces of equipment. It's a good deal all around and only cost the taxpayers the square footage of the facility and the staff salaries which pales to the $100M+ in equipment that industry has donated. And we generate income off of the contract time to upgrade and maintain the place.

From this perspective, the cluster race makes sense:

1) by getting it in the top 5 or top 10, they earn bragging rights which is essentially the best marketing you can hope for. This brings work to the university from DARPA, NASA, industry, etc. My guess is that the project was funded by NSF or DARPA.
2) even if the university can't make full use of the cluster (faculty can *always* dream up shit to run on a cluster like this) they'll rent it out to others to cover support costs.

In the mean time, the university has a resource for students to use that likely didn't impact tuition in any way, positively or negatively. Public universities have very clear separation of funds, so research grant can't be used for tuition, and tuition funds can't be used for research or buildings. It's a very controlled situation that often times causes surpluses that can't be applied to places in need, but it also makes sure that nobody is exploited (e.g. tuition being used for supercomputers instead of instructors).

Big public universities are not taxpayer supported to the degree that you might expect. Taxpayers might cover 25% or so of the operating expenses of a major research university, but the rest comes from tuition, grants, gifts, and so on. Some public universities like University of California, aren't even directly state controlled. They are essentially a contractor to the state, educating state residents in exchange for funding, but as the taxpayer funding percentage decreases, the university acts more and more privately. The California State University system is a directly funded institution, and much more akin to what you might expect.

I'm a Hokie

[Anonymous Coward]

The best thing is, they were asking for volunteers to put this thing together. I signed up for a 4 hour shift next week! I'm so excited!

Re:I'm a Hokie

[Anonymous Coward]

well you cant beat free labour, worked with the niggers so as long as you call the tutor masser you should be ok

Re:I'm a Hokie

[Anonymous Coward]

no, you're an idiot, now instead of paying someone they get it for free, great move i cant wait till you do the same for starbucks

So, do they smoke alot of crack at VT?

[Anonymous Coward]

I'm pretty sure using G5's is nowhere near having good price/performance.

Re:So, do they smoke alot of crack at VT?

[Anonymous Coward]

"pretty sure" and an a.c.

like i'm listening to you

who the fuck cares? really?

[Anonymous Coward]

i'm going to masturbate twenty times today. let's see that on /.

Actually they're not that mum

[Nefrayu]

I got the following email the other day:
Virginia Tech is in the process of building a Terascale Computing Cluster which will be housed in the Andrews Information Systems Building (AISB). For those who are interested in learning more about this project, we will host an information session on Thursday, September 4th from 11 a.m. to noon in the Donaldson Brown Hotel and Conference Center auditorium.
We look forward to seeing you there
Terry Herdman Director of Research Computing.

I'll try to remember to take notes on this and let you all know if there's anything interesting...

Re:Actually they're not that mum

[Anonymous Coward]

Great! I guess we should start an email list for people who'd like you to let them know about it, so here's mine...

just kidding

Re:Actually they're not that mum

[Johnny Mnemonic]

Yeah, I'd be interested, esp if you help with the deployment. Like, what OS are they set up to run? How are they interconnected? What are the cooling/power arrangements? etc. Email above. Thx.

PENN STATE

[Anonymous Coward]

yeA!! psu is #72!!!!!!

Not fast enough

[cperciva]

By my count, they'll have an R_peak of 8800 GFLOPS; unless they've got more efficient linpack code than anyone else, that will put them around 7th or 8th place.

Re:Not fast enough

[Anonymous Coward]

You left out a factor of 2. 2 processors * 1100 * 4 FLOP * 2 GHz = 17600 GFLOPS. Remember, the 970 has 2 FPUs each capable of 2 FLOPs/ cycle.

Re:Not fast enough

[bspath1]

Reposting my AC post:

You left out a factor of 2. 2 processors * 1100 * 4 FLOP * 2 GHz = 17600 GFLOPS. Remember, the 970 has 2 FPUs each capable of 2 FLOPs/ cycle.

Leaving out a factor of 2 in this case significantly alters the R_peak value. -Bruce.

Re:Not fast enough

[JDWTopGuy]

7th or 8th place out of 500 isn't good enough? Whatchou smokin?

Re:Not fast enough

[Anonymous Coward]

When calculating FP throughput for G4 or G5 machines, don't forget that both chips support 4-wide single precision SIMD FP including fused multiply-add operations. A 970 doing single precision FP vector work would potentially see a big improvement even over its dual scalar FPU's in throughput.

Re:Not fast enough

[Anonymous Coward]

It's all about having a good marketing bullet point for Apple. 7 or 8 is clearly not good enough for that.

Not!

[Bones3D_mac]

Of all the Macintosh rumor websites, Think Secret is, by far, one of the most reliable sites I've seen. If it wasn't, I wouldn't be investing $1,200 a year in them for their message board. Of course, if you think you can do better... ;-)

yeah, it will be fast, until....

[HomerJ]

It goes head to head with Pitt's supercomputer here. Then it will be crushed and all the time/energy will be wasted.

Re:yeah, it will be fast, until....

[Anonymous Coward]

How would it be wasted? A supercomputer is a supercomputer. Pitt can eat an ass.

Re:yeah, it will be fast, until....

[Anonymous Coward]

Tell us how fast it is all going from the big east next year.

Talk of it all over campus?

[Control-Z]

there's talk of it all over the campus

Yeah, chicks dig massive...computers.

No wait, no they don't!

Re:Talk of it all over campus?

[Anonymous Coward]

Wasn't there a former dude who became a chick who wrote some seminal work on VLSI design or something.

So there are chicks who dig compters. They just used to be guys. Sorry.

Re:Talk of it all over campus?

[baryon351]

Oh yes we damnedwell do!

Re:Talk of it all over campus?

[JDWTopGuy]

Come see mine... it's a 486, but the case is bigger than my, uh, yugo! And check out the size of the 5MB hard drive! w00t! Yeah BABY!!!!!

Re:Talk of it all over campus?

[evilviper]

Yeah, chicks dig massive...computers.

Actually, they do if it looks cool. Otherwise, no.

Re:Talk of it all over campus?

[Anonymous Coward]

Its not the size that matters, its how much you compute with it. (This is much clearer if you understand big O notation.)

Re:Talk of it all over campus?

[awx]

My girlfriend digs my VAXcluster.

Re:Talk of it all over campus?

[eXtro]

Lynn Conway, who co-wrote Introduction to VLSI Systems.

Re:Talk of it all over campus?

[dacetone]

Your girlfriend is Apache?

Memory

[rf0]

One thing against clusters rather than machines designed for the ground up is memory access. If on a n Single System Image (SSI) system is that any node can access memory of another over fast internconnects. With a cluster the memory has to be transfered over ethernet which even if using 10GB Ethernet is still a number of magnitudes lower than memory

Rus

Sounds familiar...

[jettoblack]

Right after the Sony Playstation 2 launch, there was a big shortage. Several media stories blamed it on some "unnamed" Middle East country buying them all up to power their missles and supercomputers (because, the rumor claimed, the PS2 was just so powerful).

Wonder if Apple is trying to "pull a Sony" here...

Re:Sounds familiar...

[Anonymous Coward]

My guess?
No.
While their shipping dates have been optimistic by anyone's standards, Apple has been very forthcoming on shortages in the past.

Re:Sounds familiar...

[Anonymous Coward]

shut the fuck up

Re:Sounds familiar...

[Trent05]

Pretty sure it was N. Korea. Anyone else remember otherwise!?!? Yea, the Japanese ain't too particular on a nation a couple hundred miles away having nukes and the means to deliver them. When a commie country is being asked by China to come to their senses about all that crap, I'd be pretty on the edge too. My $.02

Re:Sounds familiar...

[guile*fr]

nah... N. Korea is the >. Saddam was blamed for the shortage of PS2.

Re:Sounds familiar...

[guile*fr]

nah... N. Korea is the "ennemi du jour". Saddam was blamed for the shortage of PS2.

should have pushed preview.

Re:Sounds familiar...

[JDWTopGuy]

That's funny, AFAIK there's no flight simulator for PS2... (dumb terrorist joke)

BTW, that's a really stupid rumor.

Re:Sounds familiar...

[SiMac]

North Korea is in the Middle East?

I always knew the facts they were feeding us in school were bullshit!

Re:Sounds familiar...

[Anonymous Coward]

No, they are probably just revisiting a proud moment in Apple history, the classification of the new Powermac G4 as a weapon too dangerous to export to certain countries! Here is how Apple informed the public.

Here is more detail on this story.

Re:Sounds familiar...

[steve_bryan]

What many Intel bigots don't realize is that unlike pentiums which have their roots in calculator micro controllers, the processors from IBM are scaled down versions of a processor architecture that doesn't really need publicity stunts to enhance its reputation. For instance, were you aware that Deep Blue, the first computer to defeat a chess world champion, was a PowerPC based cluster?

The other ironic fact about your apparently clueless observation is that a super computer has been built from a cluster of PS2's at the national super computer center in Illinois. It didn't appear involve any sort of promotional discounting from Sony and it was said to provide a cost effective method of doing some quantum chromodynamic calculations. So some researchers at the national super computer institute don't seem to share your sarcastic dismissal of the PS2.

Emporium

[maukdaddy]

They're probably all just for the emporium and everyone thinks they're for a super computer ;)

Hokies will know what I mean!

Re:Emporium

[absoluthokie]

Actually... they are supposed to be getting some G5s for the Emporium :-D. These though are headed for the ISB, the Emporium isnt getting duals, probably 1.8 or 1.6ghz machines.

Re:Emporium

[CompGeek01]

LOL, Yeah man. Like we need more apples for the Emporium. Waste of space if you ask me. Hookies > UCF tomorrow baby!

Re:Emporium

[Anonymous Coward]

OMFG - I used to work as a lab monkey for VT computing services. The math emporium sucks in the worst way. I had the much better assignment at Hillcrest. In an 8 hour shift you maybe had to do 15 min of work - like unjamming a printer. One semester I played starcraft untill i almost started to hallucinate. Funny thing is two years later i was working as a Mac Genius at the Tyson's apple store....

- ljf at mac dot com

Parent is stupid

[Anonymous Coward]

You obviously know very little about distributed computing.

Latency isn't the most paramount issue, otherwise render farms and clusters wouldn't be as popular as they were today.

and lets not forget about projects like distributed.net and Seti@Home. Latency is not at all the concern for them.

Re:confirmed: I just shat all over myself

[Anonymous Coward]

Don't panic. Phone somebody you know and trust and get them to set off the fire alarm. While everybody traipses outside you can sneak off to the restroom, dispose of the "evidence" and fashion some temporary trousers out of paper towels. If anyone, upon returning to the office, should challenge you about your unconventional dress sense, shout "WHAT, ARE YOU DRUNK OR SOMETHING?" and storm out in a huff.

Well, it worked for me anyway.

In a letter to CS majors

[Anonymous Coward]

According to a CS major I know they sent out a letter earlier today asking any CS majors to help out with the system. One should also keep in mind that tech has the largest student computer lab in the world -- with almost 600 flat panel imacs (700mhz, maybe 512 of ram I don't recall).

This is being discussed

[Anonymous Coward]

on the Goatse INFO message boards

What operating system will they be using?

[Sonicboom]

The article makes no mention of the operating system that will be running on this supercomputer. I for one would like to see them get this done w/ OS X rather than use GNU/Linux.

Re:What operating system will they be using?

[Anonymous Coward]

1T'S N0T GNU/L1NUX Y0U STALLMAN W0RSH1P1NG ZEAL0T!

Re:What operating system will they be using?

[GigsVT]

Why?

Why should they waste resources on a lickable desktop environment that will never be used? Do you think all these machines will have monitors or something?

Re:What operating system will they be using?

[damiam]

You can run OSX without a GUI. It would be pretty sweet if they all did have monitors, though.

Re:What operating system will they be using?

[dbirchall]

You can run OSX without a GUI. It would be pretty sweet if they all did have monitors, though.

Yeah.

And each one would show a different timeline or possible future or past.

And there would be this guy there, called the "Architect."

Then Neo would come rushing in...

Oh, wait, I forgot, Macs get used by the good guys in movies. My bad.

Re:What operating system will they be using?

[bill_mcgonigle]

I for one would like to see them get this done w/ OS X rather than use GNU/Linux.

Do you really think Apple will donate a thousand computers for a showcase benchmark and not insist on Mac OS X?

Too bad they couldn't pull it off using PDO, but MPI is more well known, so better for adverts.

Here is proof!

[Anonymous Coward]

Got this to my vt E-mail addres a few days ago:

Hello all,

This email is to serve as invitation and notice of impending Terascale
Facility assembly assistance. For those receiving this info for the
first time know that Virginia Tech is building a top 10 supercomputer
from scratch and we need your assistance. We do have one stipulation
to volunteerism and that is you must not be a wage employee of the
university. Grad students on GTA/GRA are fine as well as others
outside the university that may wish to volunteer.

We are expecting to receive machines next week!!! Yikes! In
preparation for the assembly process, we need to get volunteers
together at the AISB (Andrews Information Systems Building), 1700 Pratt
Dr., this weekend. We are planning to have a process orientation
session start at 10:00 AM on Saturday, August 30, and last no longer
than an hour. We can give you a few more details about the project if
you show up and have not been before. :-)

There are many things that need to be covered and many new volunteers
needed. We have posted an electronic sign-up sheet for proposed shifts
at http://mm-server01.multimedia.vt.edu:9000/sign-up/ We will need
folks to sign-up as either a primary volunteer or on-call/backup person
that we can call and bring in if we are short people. We know this is
a very busy time for everyone and we want to get this done and over
with quickly so it will not affect other work that needs to be done
across campus. Once we have a definite date for the deliveries we will
send out notification to those folks that said they were available on
that day. We will have 48 hours notice of shipment, 72 hours notice of
delivery. The machines will arrive on a staggered, every other day,
schedule. Three shipments are expected for the total number of
machines.

Please forward this message to any and all listserves where we might
recruit additional volunteers. Have them respond to my email address
multimedia@vt.edu or sign-up using the above URL. It's crunch time and
we sure could use the many folks that said they could help and have yet
to pitch in. We need you now more than ever!

Thank you all,

- Jason

I get to help build it =D

[Anonymous Coward]

In my email the other day, I received this letter:

Hello all,

This email is to serve as invitation and notice of impending Terascale Facility assembly assistance. For those receiving this info for the first time know that Virginia Tech is building a top 10 supercomputer from scratch and we need your assistance. We do have one stipulation to volunteerism and that is you must not be a wage employee of the university. Grad students on GTA/GRA are fine as well as others outside the university that may wish to volunteer.

We are expecting to receive machines next week!!! Yikes! In preparation for the assembly process, we need to get volunteers together at the AISB (Andrews Information Systems Building), 1700 Pratt Dr., this weekend. We are planning to have a process orientation session start at 10:00 AM on Saturday, August 30, and last no longer than an hour. We can give you a few more details about the project if you show up and have not been before.

There are many things that need to be covered and many new volunteers needed. We have posted an electronic sign-up sheet for proposed shifts at (link deleted) We will need folks to sign-up as either a primary volunteer or on-call/backup person that we can call and bring in if we are short people. We know this is a very busy time for everyone and we want to get this done and over with quickly so it will not affect other work that needs to be done across campus. Once we have a definite date for the deliveries we will send out notification to those folks that said they were available on that day. We will have 48 hours notice of shipment, 72 hours notice of delivery. The machines will arrive on a staggered, every other day, schedule. Three shipments are expected for the total number of machines.

Orientation today was postponed, however, so I won't have more details until Wednesday =/ I'm looking forward to helping out, though.

Getting a bit ahead of ourselves

[Richard Mills]

"If it manages to complete the cluster before the cut-off date, it will score a Top 5 rank in the Linpack Top 500 Supercomputer List."

Err... I think somebody's getting a bit ahead of themselves here. =) Building parallel computing systems is complicated, and it may end up being quite a bit harder to realize the predicted performance than thought (not an uncommon occurrence). I'll believe it when they have the actual Linpack numbers.

Top 5? I dont think so

[rawgod0122]

Currently the top 5 consist of 4 machines that have a Therotical maximum speed (Rmax) the are larger then the 10TFLOPS this machine will have. Then you have to translate that into peak speed which is what matter and what this list uses to rank the machines. Peak will be a good deal less, but this mostly has to do with the way the systems are interconnected and not the machines themselves. Say what you may about the G5 but the interconnect is more important.

There is only one machine in the top 5 that this cluster could beat. The rest of the world has had 6 months to build machines too.

This should be a top 10 machine for sure. Good to see more fast machines being built every day.

Re:Top 5? I dont think so

[Anonymous Coward]

There is only one machine in the top 5 that this cluster could beat. The rest of the world has had 6 months to build machines too.

This should be a top 10 machine for sure. Good to see more fast machines being built every day.


Infiniband isn't quite up to the speed & low latency of Myrinet or Quadrics. More important, Intel just released the 1.5GHz Madison, and I guess we will see a couple of ia64-Madison entries scoring pretty high. My guess is that this machine would score somewhere in the range 12-20 by november.

Re:Top 5? I dont think so

[Anonymous Coward]

You got the specs wrong. Take a look at the CPU at arstechnica. 1100 dual G5's produce an Rpeak of 17.6TF and Infiniband has higher bandwidth (10Gbps) vs the 2 Gbps of Myrinet. Their latencies are similar. So all in all, read the specs carefuly

The Altivec stuff is the key, I'll bet.

[Above]

Several years ago I did some work on some Virginia Tech "supercomputers" (actually, baby versions of ones on campus that were the same as huge ones they leased time on elsewhere), and I think the people talking about Altivec are on track. I never knew exactly what they did, but at that time the Math, CS, and Engineering groups were working together to simulate wing designs for the YF-22 jet figher prototype. Since I was more of a "sysadmin" (althoug h with a math and CS background) I ignored most of what was going on, but one thing I can tell you was vectors, vectors, and more vectors. The vector is king. It's an assumption, but I'll bet they are still working on similar type studies, and if built, this will be just the beast for it.

Re:The Altivec stuff is the key, I'll bet.

[Anonymous Coward]

I keep noticing that it's only the astroturfing Apple shills who keep bringing up 'AltiVec' in this discussion.

Is there some sort of bonus if you can work AltiVec into your comments this week?

Re:The Altivec stuff is the key, I'll bet.

[Ninja Programmer]

I ignored most of what was going on, but one thing I can tell you was vectors, vectors, and more vectors.

For the Linpack test to get on the top500 computers? That would be "64-bit FP vectors, 64-bit FP vectors, 64-bit FP vectors" which AltiVec cannot do.

Re:The Altivec stuff is the key, I'll bet.

[Anonymous Coward]

Actually, you could fiddle the numbers and turn a 4 single precision instruction into a 2 double precision instruction. But your numerical algorithms have to be tight, and it would undoubtably slow things down. I put more confidence in the 2 independent FPU theory.

Re:The Altivec stuff is the key, I'll bet.

[Anonymous Coward]

The guy was talking about what he believes the cluster may be used for, not the sodding Linpack test.

I don't think...

[JeffTL]

...it is in my power to imagine a Beowulf, or indeed any large-scale cluster, of G5's...I'm overloaded.

Fucking A

[waldoj]

Damn, I've been at this school for a week and I haven't found a single redeeming value. Finally, a cause to hang in there for the next couple of years.

-Waldo Jaquith

Poor choice on Apple's part

[SuperBanana]

Ok, so let's get this straight- Apple just screwed 500 customers for the sake of getting their name on some supercomputer top-500 list? I know if I had ordered my G5 and found out my machine was a month further away than I thought it was, I'd be spitting mad; the only people who have that kind of dough are people who really need the speed for business purposes. Not to mention this could be the straw that breaks the camel's back WRT Apple executive credibility with analysts- Steve said the G5's would be shipping by the end of August, and this cute little Virginia Tech deal will push that up almost a month.

I'm a diehard mac fanatic, have been for years- I've bought something like 6-7 Macintoshes over the last 10+ years, and I'm about to buy another in a day or two...but sometimes Apple management just completely stuffs its head up its ass.

As for Virginia Tech, this couldn't possibly be an efficient choice for a cluster; AMD Opertons in rackmount configurations maybe- but not G5's. For about $1k less than the dual G5, you can get a dual opteron system similarly spec'd out, and you can fit a couple of them in the same amount of space as one G5. No, this sounds like a publicity stunt more than anything else...

Re:Poor choice on Apple's part

[Glonoinha]

-I'm about to buy another in a day or two...

You mean you are about to order another in a day or two ... there is a pretty big gap between asking for one and actually getting one. Tell you what, let me finish this gig I got happening in Virginia and then we can start dealing with the customers that want to order G5 machines in the onsies - twosies quantities.

Sincerely,
Steve.
sjobs@apple.com

PS - Seven Mac's in ten years isn't hardcore. 1100 units ordered on one purchase order before they even ship (August, 2003) ... now THAT is hardcore.

Re:Poor choice on Apple's part

[Anonymous Coward]

Not a single vendor anywhere could provide the numbers of 64 bit amd or intel hardware required. Apple came through.

Re:Poor choice on Apple's part

[runenfool]

You got a link for dual Opterons for 1500 bucks with all the goodies in the G5? If yes, Id love to see it.

(seriously - I can take out a loan :) )

Re:Poor choice on Apple's part

[SuperBanana]

PS - Seven Mac's in ten years isn't hardcore. 1100 units ordered on one purchase order before they even ship (August, 2003) ... now THAT is hardcore.

Yes, but the difference is, to buy one I had to mow lawns for an entire summer, and my parents chipped in the other half.

What's the most you had to do, remember where your pen was, and shlep the PO down to Purchasing?

Re:Poor choice on Apple's part

[benh57]

And that opteron system will use 3x the power and heat. Universities don't spend millions on a 'non efficient choice'. Smart people did the math, and they had facts - which you don't.

Re:Poor choice on Apple's part

[Glonoinha]

It was a joke, a parody. I'm not really Steve Jobs and I don't work at Apple.

The most I have ever had to do? Let me think back.

I was in college at the time, working a full time load my last year in school. 1989, as I recall. Working on my Senior Project I knew that if I didn't have unlimited access to a machine with a compiler (Borland's Pascal, for the record) I was not going to be able to finish it on time. The college had a computer room but generally it closed at 9pm or something insanely early like that, and was generally full of other students - so I knew I needed a machine.

I was making $5 an hour, working 40 hours a week (in addition to my full load at school)at a job 30 miles away. No parents, I was living on my own with a room-mate. I ended up selling my pistol (a nice S&W 6906 if it matters), the only thing I owned of any value whatsoever besides my motorcycle, and bought a 386sx-16 with 1M of RAM, an 8-bit VGA card with 256k of memory, both floppies, no hard drive, a 1200 baud modem, and a monochrome 12" vga monitor. I had some money saved up to add a dot matrix printer, and for a full semester programmed on that. Every day that semester I made every life decision, be it clothes, food, entertainment, liquor, caffeine, air conditioning / heating (I went the entire year in S. Texas without air conditioning, temperatures running 100+ for weeks at a time) to save money and when the next semester started I bought a used 40M hard drive, a 14" color VGA monitor, and another meg of memory.

By the time I was done the machine had cost me roughly $1,600 - earned at my $5 an hour job over the course of a year after paying for all my living expenses.

So yea, I hear ya.

Of course my last machine, the one I bought in May, I didn't think twice about. A new Dell P4/2.4 machine with a gigabit NIC, 128M of RAM, 40G of fast drive space for $300 delivered to my door took all of about three minutes on the web to configure and another two to type in my credit card number. (Where were these supercomputer class machines for $300 when I was a poor college kid?) Disclaimer - I have since added 512M of RAM, an SB Audigy and a Geforce 440mx w/ 64M to make the machine a well rounded development / game environment, bringing the cost to $500 total.

Just curious - what is the easiest you have had it when buying a machine?

Re:Poor choice on Apple's part

[Glonoinha]

-Universities don't spend millions on a 'non efficient choice'.

Yes they do. Look at the football teams of most colleges.

I think you vastly overestimate the cognitive prowess if the administrative personnel at your college. Statistically speaking, 85% of the people making important decisions were C students and scored less than 600 on the math portion of their SAT (of a 800 possible max.)

I also imagine you haven't actually sat in on any high level focus and direction meetings at the CxO level or even the Director level. Most of the time these people are not NASA quality rocket scientists and engineers, and often decisions are made not on the technical merits of a particular issue but on how charismatic the individual proponents of that solution are - presented correctly in a smooth manner, serving three day old turkey chopped up into little pieces in turkey tetrazzini is something college level administrators will consider, and possibly even do.

My advice - when you are in a Director level focus and direction meeting as a technical resource, resist the urge to dream up the stupidest, lamest idea you can and present it as if it was a wonderful idea because they honestly don't know the difference and you will be pulling coax and converting your system back to Token Ring-16, unable to convince them that you were just kidding. No shit.

AltiVec

[charnov]

While the AltiVec unit is very impressive, The SSE2 unit on the P4 or the Opteron would have nearly the same performance and cost a whole heck of a lot less (I am betting if this rumor is true at all, then Apple has given the units to the school).

What I am wondering is, what OS is this cluster going to run? I mean, have the BSD folks figured out how to scale? No chance it will be OS X...maybe AIX?

Re:AltiVec

[Space Coyote]

While the AltiVec unit is very impressive, The SSE2 unit on the P4 or the Opteron would have nearly the same performance and cost a whole heck of a lot less (I am betting if this rumor is true at all, then Apple has given the units to the school).

Real world numbers don't bear this out. Check out the Photoshop and other application performance numbers for this. The gcc version used by the SPEC benchmarks used by Apple didn't even take advantage of AltiVec. When accounted for, and any institution making such a purchase would definitely have considered this, the AltiVec-enabled PowerPC chips totally spank x86 and others in number crunching tasks.

What I am wondering is, what OS is this cluster going to run? I mean, have the BSD folks figured out how to scale? No chance it will be OS X...maybe AIX?

An OS doesn't need to 'scale' to be a member of a cluster. It just needs to run the code locally and send the result back to the cluster master node.

Re:AltiVec

[Anonymous Coward]

SSE2 and Altivec are not identical, so there's still areas where a G4/G5 would (and does) thrash an intel-anything to pieces - the same can happen with other areas where the Pentium would kill the PPC.

One example is in RC5-72 crunching. My plain old consumer eMac can crunch 10.47 million keys per second using altivec. It's not even an optimal architecture for getting all the data through to the altivec unit (133mhz bus etc). The fastest x86 based box doesn't come -near- that. (see http://n0cgi.distributed.net/speed/ for stats)

That's one extreme case, and possible the most extreme one. Only knowing the final purpose for the VT cluster would reveal just how sensible picking PPC machines is.

Re:AltiVec

[SirDrinksAlot]

MacOSX is a fabulous OS for multiprocssing and clustering. Theres no reason not to have it on there at all.

Apple has their own clustering software and theres also software for OSX available from the AppleSeed project (google for it if you like)

Theres also the ability to run CLI only on OSX so that can free up processor time too.

Re:AltiVec

[discstickers]

The SSE2 unit on the P4 or the Opteron would have nearly the same performance and cost a whole heck of a lot less.

Uh, no. 2 years ago, my roommate and I were both running the distributed.net client. I have a 500 Mhz Powerbook G4 (100Mhz bus). He had a 1.4GHz P4 with rambus RAM. I got 4Million keys/sec. He got 2MKeys/sec.

So clock for clock, my machine was nearly 4 times faster.

Re:AltiVec

[Anonymous Coward]

I don't think so. see here

Re:AltiVec

[tealover]

Uh, no. 2 years ago, my roommate...

I thought you guys preferred the term "life partner" these days.

Re:AltiVec

[Josh Booth]

Of course it is. Mac OS X IS UNIX-LIKE. FreeBSD, more precisely. Anything you can do with the average *nix you can do with OS X. I doubt that Apple reimplemented anything. They probably just made current capabilities pretty.

Theres also the ability to run CLI only on OSX so that can free up processor time too.

You speak as if CLI is a feature unique to OS X. *nix is very modular, and anything can be enabled or disabled, depending on what you need. So, you are describing a UNIX-like system.

Re:AltiVec

[Anonymous Coward]

LOL! Oh that's a killer.

Re:AltiVec

[Anonymous Coward]

SSE2 is *very* slow compared to Altivec. They're not in the same ballpark. Apple (or more correctly, Motorola) did an absolutely first-rate job with Altivec.

Re:AltiVec

[JDWTopGuy]

No kidding. My 667Mhz Powerbook G4 gets 6.8 MKps, my friend tried it on his 2.53Ghz P4 and got 3.4 MKps, and my 1.4Ghz Athlon XP 1600+ gets aroung 4.5 MKps. I calculate a Dual 2Ghz G5 around 40 MKps, probably more. Imagine them running distributed.net on this bad*** cluster!!!! YOW baby!

Re:AltiVec

[Wesley Felter]

Too bad distributed.net is completely unrepresentative of scientific applications (or any real applications for that matter).

Re:AltiVec

[Anonymous Coward]

have the BSD folks figured out how to scale

It's an cluster of dual processor SMPs doing compute bound tasks. And if you had half a clue you'd be asking the Mach people if the thread scheduler works, the BSD stuff in OS X sits atop it. Oh and its only been that way for something like 15 years since NeXTSTEP was started.

Re:AltiVec

[discstickers]

While that may be true, that has no bearing on the fact that AltiVec is very powerful and can be used for scientific applications. Blast scores here (pdf). Also, check out this article from O'Reilly.

Re:AltiVec

[Anonymous Coward]

Isn't AltiVec just the new whizzy-word at Apple?

They were using RISC as their wizzy-word a decade ago, and they seem to go through a few of them every few years.

Re:AltiVec

[anthonyrcalgary]

Isn't IBM releasing xlc soon?

hmmm...

Re:AltiVec

[Anonymous Coward]

IIRC Altivec had some swizzle instructions that weren't included in SSE or 3DNow!.

However PNI (Prescott New Instructions, SSE3?) may fill in the missing instructions that altivec has.

If what you are doing depends on those instructions the G4/G5 will run faster than SSE/3DNow!. However if you're doing MADD (multiply+add, key instruction for vector/matrix math) SSE and 3DNow! handle that just fine and Altivec has no advantage. At that point it comes down to how many instructions you can execute per cycle, how many cycles per second you can execute instructions, and how quick your memory is. :)

It would seem this is great for the P4, and I guess the G5.

Re:AltiVec

[drinkypoo]

While the AltiVec unit is very impressive, The SSE2 unit on the P4 or the Opteron would have nearly the same performance and cost a whole heck of a lot less

Real world numbers don't bear this out.

If you're talking about buying the Macs, and buying pre-assembled PCs, and comparing their power dollar for dollar, then you're right, the macs are more powerful. If you're talking about building clones from commodity hardware, which is simply not an option when it comes to macs, then I think that you are wrong, sir.

Assuming you're willing to buy surplus machines, for the price of a base-model 1.6GHz G4 ($1999) I can buy three 2.53GHz P4s, and still have money left over to pay for the difference in shipping costs between one PC and three PCs. The HP P4s probably only have hundred megabit ethernet, but they also have IEEE1394 so you could make a number of 400Mbps links between nodes in order to improve throughput, if you so desired.

In addition, each of the P4s has a CDRW, a DVD-ROM, 512MB and not 256MB ram, an 80GB disk (three for the price of one!) If you only bought two of them, you could afford gigabit nics, and I'm guessing two P4 2.53GHz CPUs will beat one G5 1.6GHz.

For the money, you can't beat Intel. As far as I can tell, either they're getting the machines at or near cost from Apple, or they're concerned about space and power considerations, which are a real issue. I should not have to remind anyone that power is a consideration twice; supplying it, and removing the waste heat. But one of the nice things about a cluster is that you can replace it in bits, and if you're willing to compile and/or optimize for different platforms, they need not even be the same architecture.

An OS doesn't need to 'scale' to be a member of a cluster. It just needs to run the code locally and send the result back to the cluster master node.

This brings me to my next point. A cluster isn't a cluster without Miracle Whip. Er, I mean, without process relocation, if not thread relocation. I know, I know, it's still more or less a cluster, I've used DQS for example, but it's not transparent in the way that OpenMosix is, for example (which is still x86-only, right?)

Anyway if I were building a cluster today I'd order up all the refurb'd HP P4s and Athlon XPs I could afford from geeks.com since they have such nice prices, then I'd load them with Linux and OpenMosix, or if what I was doing could be handled in a job-submission fashion, I'd load them with some x86 Unix or Unix clone (Still probably Linux, but maybe FreeBSD) and use DQS to handle job processing. But if I were doing this it would probably just be a handful of machines, and it wouldn't be worth stressing over power requirements and heat dissipation.

Re:AltiVec

[Anonymous Coward]

I don't think so. see here

Hmmm... I think *you* should check the responses that your very credible AC post generated. They explain why it is quite possible to achive the speeds that these guys have reported.

Re:AltiVec

[valdis]

soon as in the past tense? It was a Slashdot article a few days ago

http://apple.slashdot.org/article.pl?sid=03/08/2 8/ 1746213&mode=thread

Re:AltiVec

[sevenofnine]

Yeah right, so what your saying is that they should build their own cluster by putting together 1100 units of grey boxes?
In your "calculations" i think you forgot to add the added manhours for doing this...
and thats not even thinking about the order / failure rate...
buy mem for 1100 machines is bound to have 1, 2 or more (properbly the ladder) mem blocks not working...
And im sure there is alot more reasons to this that you ( and i) missed...

Re:AltiVec

[Anonymous Coward]

Too bad they locked it down to a dying arcitecture.

Re:AltiVec

[drinkypoo]

They're already going to put together 1100 units of systems in a cluster. You'd need to use 3300 of the HP systems I noted in order to get the same performance, which is probably why they aren't doing it that way, but it would still be around the same price. Perhaps they require 64--bitness for something they're doing.

Meanwhile, refurbished systems are tested, so I don't know what you're bitching about, or how you got modded insightful. Doubtless you realized you couldn't bitch and moan at me about power consumption and space utilization since I mentioned it already.

Actually I don't think those 2.53s are even refurbs, I think they're discontinued overstock, because they're just too new to actually have been deployed and then withdrawn. They come with warranties, in factory packaging, so the only real issue is the price/power consumption+space (*space?) ratio.

Re:AltiVec

[fm6]

So clock for clock, my machine was nearly 4 times faster.

Yes, and clock for clock, your machine is 4 times as pretty. They're equally irrellevent.

As this very experiment ought to tell you, CPU clock speed is only one factor in many determining how quickly the CPU will do some particular thing. It is impressive (but not suprising) that a 500 Mhz G4 does basic number crunching twice as fast as a 1.4GHz P4. But a 1.4GHz G4 wouldn't necessarily crunch 2.8 times as many keys as your system does. Probably not.

Re:AltiVec

[discstickers]

Actually, a 1.25GHz G4 does 13Mkey/sec. Now, I think that is a dual processor score. So each processor is doing 6.5Mkeys/sec. 2.8*2MKeys/sec = 5.6MKeys/sec. So the newer G4s are actually more efficient (remarkable considering the bus multipliers involved).

Re:AltiVec

[fm6]

Either way, there isn't a linear relationship between processor speed and ability to crunch keys. Bottom line: Yes, a G4 is much, much, much faster than an equivalent P4. But there's no obvious quantification of "much, much, much".

I'm not an expert on this stuff, but I've worked with engineers who are. One thing they hate is simplistic benchmarks. Mainly because you can always find one that "proves" your platform is better than the others.

Re:AltiVec

[cosmo7]

Isn't AltiVec just the new whizzy-word at Apple?

They were using RISC as their wizzy-word a decade ago, and they seem to go through a few of them every few years.

Wow! George W Bush reads /.!

Hoax

[Anonymous Coward]

The cluster has nothing to do with Apple. They were contemplating using PPC970 CPUs (basically becasue they're cheaper than the Power4's), but AFAIK there were never plans to use complete G5 computers, or anything else made by Apple. And I doubt IBM would take CPUs from Apple to give them to V.Tech. The delay is mot likely due to performance issues; they're probably fine-tuning the OS so it's competitive with x86 / Windows systems.

Re:Hoax

[Anonymous Coward]

The cluster has nothing to do with Apple. They were contemplating using PPC970 CPUs (basically becasue they're cheaper than the Power4's), but AFAIK there were never plans to use complete G5 computers, or anything else made by Apple. And I doubt IBM would take CPUs from Apple to give them to V.Tech. The delay is mot likely due to performance issues; they're probably fine-tuning the OS so it's competitive with x86 / Windows systems.

This actually makes a LOT of sense in my experience. For a cluster of this size, the continous administration and replacing failed hardware is going to cost way more than the initial aquisition of hardware.

It would be hell to administrate 1100 off-the-shelf big boxes where you have to unplug each box manually if there is a problem, and you can't start/stop the hardware remotely.

Unless they are getting some yet unknown XServeG5, I'd say it is much more likely this machine will be based on IBM PPC970 blades.

Re:Hoax

[jarrell]

Uh, no. That vast farm of already-cabled empty racks in the machine room are waiting on G5s to be dropped into them. Much thought has gone into airflow and cooling. If a as yet unannounced G5 xserve becomes available in the future, the racks are well positioned to swapping them in. Assuming the same form factor as the current xserve, they can handle 5 xserves in the space where three g5 are now.

Re:Hoax

[Gogo Dodo]

Ripping the processors out? To do what with them? You can't just stuff them into any old motherboard.

OS X is already competitive with Windows.

How did this get moderated Insightful/Interesting? The moderators are nuts. I hope the meta-moderators correct this. This is a Troll/Flamebait all the way.

How would seti@home rank?

[Anonymous Coward]

Or folding@home? In a way they do satisfy the criteria for a cluster, so on average where would they be in the ranking?

XServe?

[diesel66]

This won't help sell their clustering hardware:

http://www.apple.com/server/clustering.html

Another brilliant idea, inspired by Apple

[Anonymous Coward]

With 1100 machines in the cluster, there must be _at least_ 2200 DIMMs. Since these must be 400MHz (PC3200) DDR, they can't be on a large 0.15 micron DRAM process, but most likely between 0.11 and 0.13u.

Who cares?

APPLE G5'S DO NOT SUPPORT ECC.

The random bit error rate for 2200 DIMMs with 0.13u cells is roughly one '1' bit dropped to '0' every 9 hours. In other words: good luck getting any reliable, large-scale computation done with this cluster. (And I do mean "good luck" - they might get a run of two or three days without any problems once in a while.)

Now if only Apple would support PC3200 ECC DIMMS, which certainly do exist:

http://www.intel.com/technology/memory/ddr/valid /d imm_results.htm

this cluster might be a bit more useful for real work.

Re:Another brilliant idea, inspired by Apple

[Anonymous Coward]

You'll be even happier when you find out your federal tax dollars funded the project...

Re:Another brilliant idea, inspired by Apple

[IM6100]

I remember back when it was a marketing bullet point for Apple people to say 'we don't waste an extra chip on the SIMMs for that useless parity bit.'

Of course, that was a long time ago in the days of 30 pin SIMMs. But MacOS back then was crash-prone enough that it probably would have been a waste to worry about hardware issues like parity in the memory bank.

Re:Another brilliant idea, inspired by Apple

[MoneyT]

Chances are they didn't. Generaly at the university level, a project of this magnitude is funded by grants or private investment.

Re:Another brilliant idea, inspired by Apple

[Anonymous Coward]

...and most of the grants come from federal research funds. According to a comment above, this might be funded by DARPA, which is as we all know a federal agency.

Re:Another brilliant idea, inspired by Apple

[dbirchall]

With 1,100 machines in the cluster, you'll probably be running into something's mean time between failures pretty darn often, whether it's memory getting a bit wrong, or one of the 9,900(!) fans needing replacing. :)

But... a cluster should be redundant enough to withstand that sort of minor inconvenience and go on functioning without the errant node while it gets fixed, reboots or whatever.

I'll admit that building something smart enough to say "Node 206, you have a memory error. Bad G5, no donut!" is beyond the scope of my understanding.

Re:Another brilliant idea, inspired by Apple

[TheRaven64]

Node 206, you have a memory error. Bad G5, no donut!

Mmm. Donut-powered G5...

Re:Another brilliant idea, inspired by Apple

[asuzuki]

You could probably use error-correcting codes (on the software level) or other means to increase redundancy and minimise the probability of an error.
I mean if this was really an issue, they would have thought of it before starting a multi-million dollar project.

Re:Another brilliant idea, inspired by Apple

[Anonymous Coward]

You could probably use error-correcting codes (on the software level) or other means to increase redundancy and minimise the probability of an error.

Easy in theory, but way too slow in practice. When you can pump 6Gb/second from memory it would take more than the peak performance of the machine just to calculate the parity.

I mean if this was really an issue, they would have thought of it before starting a multi-million dollar project.

I'm not sure :-) Even normal memory has parity, and will detect an error. Since it can't correct it that node will hang (if it is in the OS) or at least the job will die. The advantage with ECC memory is that single bit errors are corrected, and two bit errors detected.

Single bit errors are pretty common in our logs, but they will only be a source of irritation on the Apple cluster when jobs die. The bad thing is dual bit errors which cannot be detected at all without ECC - instead they silently corrupt your data. I've seen that being detected by our ECC once or twice...

Re:Another brilliant idea, inspired by Apple

[laird]

Actually, this has been done before. Back when I worked at Thinking Machines, the CM-5 had an entire network devoted to detecting hardware failures and remapping the CPU's. It was way cool -- every chip had a self-test and an a connection to the test network, and if there was a failure the OS would swap the task on that CPU to another CPU and the router would map all communications so that the new CPU looked just like the old one. Some of the compute jobs that people run on supercomputers are _huge_ (some run for months) making it very important that errors are handled properly. The machine could also swap all RAM to disk absurdly quickly, so you could checkpoint your calculations so that if there was a complete system failure (e.g. power) you could pick up from the last checkpoint instead of starting from the beginning.

MOD PARENT DOWN

[Anonymous Coward]

My plain old consumer eMac can crunch 10.47 million keys per second using altivec

LOL mac zealot. didn't think that one through did you. an eMac doesn't go any higher than 1GHz. If you think a 1GHz machine has any hope of doing over 10 million rc5 keys you're really trying to fool yourself.

Re:MOD PARENT DOWN

[class_A]

ooops, yuo = own3d!!!!
http://n0cgi.distributed.net/speed/query.php?cputy pe=99&cpumhz=1000&recordid=1&contest=rc572&multi=0
Power PC 7450/7455 G4 1000 MacOS X 10.2 2.9005 RC5-72 10,594,666.00

Re:MOD PARENT DOWN

[Anonymous Coward]

LOL another mac zealot! look again at all of them in the first column REPORTED AS ERROR.

Re:MOD PARENT DOWN

[ZackSchil]

That first column says "Report as Error" It's a link to report the number as an error if it is unusually high compared to its neighbors, not to state that the actual figure is an error. Click the link, it will explain the system to you. I must be off because being a homosexual, a Macintosh user, and a priest takes a lot of my time! At least the slashdot trolls seem to think so! Toodles!!

Re:MOD PARENT DOWN

[Anonymous Coward]

wow. I was looking at a lifestyle change and thought I'd see if I could become a complete and utter shithead. I thought to myself "hey, you could do really good at this" but looking at your post I see I'm beat before I even started.

Re:MOD PARENT DOWN

[SirDrinksAlot]

Read this at then go back to school.

Re:MOD PARENT DOWN

[shiftyphil]

Or you could just check the site yourself and shut the hell up. This test is Altivec's biggest win, the G4 absolutely embarrasses systems with three times the clock speed.

Re:MOD PARENT DOWN

[Lars T.]

Erhm. He was replying to an AC claiming "LOL mac zealot. didn't think that one through did you. an eMac doesn't go any higher than 1GHz. If you think a 1GHz machine has any hope of doing over 10 million rc5 keys you're really trying to fool yourself."

Re:MOD PARENT DOWN

[ZackSchil]

Thank you. The slashdot mod system kinda makes replying to things a moving target. When in doubt, click the parent link.

Making... oh... ONE good thing about the Empo.

[VT_hawkeye]

Well, maybe 2. The chairs are pretty comfortable.

Kinda curious why the Empo needs them, especially in the budget crunch. But I shouldn't complain (especially since I'm now an alumnus), I'll just watch University Surplus and score a used G4 from there for cheap once they retire them in favor of the G5s. (WOOHOO!)

Re:Making... oh... ONE good thing about the Empo.

[Anonymous Coward]

I really can't see the Emporium getting more than a couple of G5s. They just replaced all the computers with new iMacs last year.

Conspiracy Theory

[Anonymous Coward]

1100 seems too close to 1024. Perhaps the cluster is only 1024 computers and they said 1100 so they can get some Duals for other purposes (faculty, maybe even students) even faster.

American Supremacy in Supercomputers

[reporter]

After the introduction of the supercomputer called "Earth Simulator" by NEC, many Americans went into paranoid mode. They feared that the Japanese "once again" had taken the lead in a crucial technology.

American fears are unfounded. Numerous universities like Virginia Tech have trained a generation of American (not foreign) students in building the finest supercomputers. MIT, Carnegie Mellon University (CMU), and Virginia Tech (to name just a few) have launched large-scale research projects staffed by top American graduate students. Their work became the foundation of several generations of multiprocessors.

By contrast, very few (if any) Japanese universities conduct large-scale research projects to build high-performance supercomputers. The Japanese government has tended to avoid funding this kind of research. Worse, there is little collaboration between industry and academia in Japan. Yet, precisely this kind of collaboration is needed for such large-scale projects: e.g. Virginia Tech is enlisting the help of Apple computer.

American companies lead by scientists trained at MIT and CMU could easily design a computer that outperforms the Earth Simulator. These companies simply have chosen to not do so because there is far more profits to be garnered by building commercial supercomputers geared for database transactions. In fact, the highest-performance commercial supercomputers nearly all come from the United States of America (IBM).

The 21st century remains Pax Americana, not Pax Asia. The hordes of immigrants trying to get the hell out of Asia and into the USA underscores this fact.

... from the desk of the reporter

I can confirm this

[mattgreen]

I know several people working directly on it and have been offered a chance to work on it (both set up and afterwards). I think I will take them up on the offer just to see it in action.

Won't someone please think of the social sciences?

[Decaffeinated Jedi]

Ah... now all the budget cuts to Virginia Tech's Department of Political Science make sense. ;)

DecafJedi

I, for one...

[kdsolutions]

... welcome our... oh wait... I hate MACs...

.\

Re:I, for one...

[scrod]

... welcome our... oh wait... I hate MACs...

Damn, ethernet controllers must really piss you off then, huh?

Re:I, for one...

[Anonymous Coward]

Yeah, Mandatory Access Control can really kill the usability of a system.

Oh wait...did you mean Message Authentication Codes?

Or maybe you were referring to Project MAC?

(Ok, someone else alredy came up with the fourth obvious alternative)

Re:I, for one...

[Ninja Programmer]

Damn, ethernet controllers must really piss you off then, huh?

How so? Xerox PARC had all sorts of good inventions that are worth while and survive to this day.

Re:I, for one...

[Anonymous Coward]

The OP *did* claim to "hate MACs".

so no G5 Xserves soon?

[johnpaul191]

i would take this story to imply that a G5 powered Xserve is not going to be shipping anytime soon..... the Xserve is made to cluster and run in situations like this. i guess the rumor sites can speculate if it's G5 parts available or some other holdup on a G5 Xserve.
unless there is some reason the desktops are better for this project that i did not pick up on?

as for the above question about Macs.... depending on what they want to really do with this, Altivec is really efficient for some computations. all flame wars aside there have always been people clustering Macs for certain uses. i do not know how much of it was user preference or the software they wanted to run or the simplicity of getting the cluster running.
it is supposedly VERY simple to cluster Macs. there was a story on /. a year or so ago about a group that went from building a rack and unboxing their G4s to a running cluster in part of a day. i really don't remember the specifics but i think it was something like 30 G4s? i would guess the G5 is not that much harder... and they seem to have Apple helping. maybe they hooked up the optical cards from the Xserve...... we'll see i guess.

Re:so no G5 Xserves soon?

[jo_ham]

Have you seen the size of the heat sinks on the G5? I saw one in an Apple store today and was very impressed with the engineering of the whole machine.

The heatsink is a large oblong about 5"x4"x6" with a thin grille like construction. It's just too big to go in the 1U Xserve. Give them some time to work on designing it to fit though. The G5 is an ideal CPU for the Xserve as you say.

Re:so no G5 Xserves soon?

[drinkypoo]

The G5 is an ideal CPU for the Xserve? What hoohaa. G3s make fine servers for most tasks, a G4 is generally overkill, unless you are demanding a great deal of CPU, which is just not necessary for most server tasks. Usually all you need is decent bus bandwidth (the G4 provides this already) and a good network interface, plus plenty of system memory for caching. The G5 has little advantage over the G4 in terms of a file server, web server, DNS server, et cetera; it is only substantially more useful as a compute server, because the CPU is so much more powerful.

Re:so no G5 Xserves soon?

[Lars T.]

So why do PC makers insist of using P4 Xeons and Opterons in their servers, if a P2 or K3 would suffice?

Re:so no G5 Xserves soon?

[Anonymous Coward]

Umm... no. The G5 and the surrounding architecture have a tremendous amount more bandwidth than the G4. Admittedly the G5 would also be much better than a G4 as a computational node but even for simple, heavy load file serving it almost certainly has much higher capacity than a G4 system.

Re:so no G5 Xserves soon?

[jo_ham]

So walk up to a person in the street and ask them if they want a 3Ghz P4 or a 900Mhz G3 in their brand new expensive computer.

I know people (especially those considering an Xserve) should understand the differences between processors, but one of the criticisms levelled at Apple about the Xserve is that the CPU is underpowered - the G5 will fit there if for nothing else than pure horsepower.

Re:so no G5 Xserves soon?

[bill_mcgonigle]

They need the 90nm process first, to keep the machines cool enough. Rumored to start being available in the November/December timeframe.

Of course IBM just laid off 600 workers at that plant, IIRC.

Re:so no G5 Xserves soon?

[johnpaul191]

I guess it depends on what you are running the machines for. If you are running an intranet or webserver, then i am sure a G3 would be great... and cheaperish? yes, an AMD chip with linux is way cheaper, but......... we're talking Apples here.

i think Apple is trying to flex it's processor's muscles with the ability to link some Xserves (be it the current G4 or the future G5) and generate some real work out of the machine. A few people have talked about running them in rendering farms or for doing massive medical/scientific computations. Taking something like rendering (vector intensive) that is often done by clusters now and running software optimized for the G4/G5 (Altivec) then you will get some impressive results. The question of price/performance matters more in a world like that than desktops when people fight over who's machine can do a photoshop filter 1.8 seconds faster.

If you look at the G4 desktops that were for sale when the G5 rolled out and the prices/specs for the G5... it seems like they are actually pretty close. The G5 starts at $2000 but every one has a superdrive, as well as a few other things that seem to keep it in the same pricing world. They did not start as cheap as the G4, but they did not roll out a cheaper thrifty model (maybe when chips production is up to speed or the current speeds are upgraded?). Instead they will be keeping the G4 towers for a while (Schiller said they will make them as long as people buy them). That makes me wonder if they can eventually make a G5 Xserve for a close price match to the current G4. If not, it would be fun to have a thrifty G3 rackmount for easier server tasks (won't ever happen).

Re:so no G5 Xserves soon?

[drinkypoo]

Because people will give them the money. The REAL law of supply and demand is, if you demand something, someone will supply it :P Of course, you demand it with money. Money is the only thing that corporations listen to. If people were demanding "outdated" processors in their servers, in order to save money where it wasn't necessary to spend it (unless you're compiling the OS over and over again, as you do with gentoo which is really a statement about what is possible when you have a ton of CPU lying around) then people would sell servers like that. However, most people who know enough to know they don't need to blow their cash on the latest greatest just buy the shit and make their own servers.

Don't get me wrong, for certain server-related tasks (databases come to mind) CPU is necessary, but for fileservers especially, it's mostly irrelevant. You have to be able to shovel the data; The G4 is a good little data shoveler.

Bragging Rights

[Michael_Burton]

I've been tempted to order a dual G5. I've resisted the temptation by realizing that my only real reason for wanting it would be to awe friends and co-workers. Pretty shallow. I was ashamed.

What a surprise to find that the folks who buy multi-million dollar supercomputers seek some of the same shallow satisfaction that moves me--bragging rights.

Still, if a single order for 1100 units causes significant delays filling orders for other customers, Apple must not have been expecting to sell many of these things. Maybe I should place an order just to help out.

Re:Bragging Rights

[evilviper]

that my only real reason for wanting it would be to awe friends and co-workers.

How about getting a stable and reliable system, that outputs less than half the heat (and use about half the electricity) than x86 counterparts?

Frankly, I have to wonder why this order was given to Apple, instead of IBM. Unless you really want to run MacOS, what's the point in buying from Apple?

Re:Bragging Rights

[Johnny Mnemonic]

Frankly, I have to wonder why this order was given to Apple, instead of IBM. Unless you really want to run MacOS, what's the point in buying from Apple?

That's a pretty good point. Can you actually buy a 970-based blade from IBM, and have a cluster in place in time to make the list? Including the Altivec parts, which presumably are part of the attraction? If the answer to both of those is "yes", I can't see why either. The benefits of OS X would be totally lost in a cluster, which would probably be running Darwin (or a stripped for of OS X, at the least) anyways.

I signed up to help build it

[tansey]

About a week ago, VT CS (and I assume Comp Eng) majors got an email telling us about it (though the email claimed it was top 10, not 5, in the world). It also invited us to volunteer to help build it. Needless to say, I signed up ASAP.

They're planning on having 3 4-hour shifts each day. They don't really know when they're going to receive the chips, just that they will get 72 hrs of warning before arrival.

The 5th fastest machine is already faster...

[Anonymous Coward]

The world isn't standing still - this machine *might* have scored around position 7-8 on the LAST list (8800 peak, but Infiniband is slower than Myrinet/Quadrics). PNL just presented their new IA64 machine that reaches 11.8Tflop:

http://www.pnl.gov/news/2003/03-33.htm

There will probably be several other machines presented before november, so my guess is the VT computer might be number 10 to 15.

TS Message Boards

[Anonymous Coward]

And as a frequent lurker there, I thank you sir.

*doffs hat, resumes lurking*

It's true.

[Mekabyte]

For those unbelievers, here's a little proof and maybe a bit more

Here's an official word (search for Teraflop).

Also, here's the original e-mail that went out (a month ago) They never mentioned Apple though:

> Date: Mon, 28 Jul 2003 17:36:46 -0400
> From: Jason Lockhart <multimedia@vt.edu>
> Subject: Terascale Assembly Assistance
>
> Hello all,
>
> As you may know the College of Engineering in conjunction with the
> university Information Systems and Computing organization are
> building
> a 10 TeraOp, 1100 node supercomputing cluster. We are in need of
> volunteers to assist in three areas of assembly; cabling, RAM and PCI
> card install, and machine racking. We would like to get as many
> volunteers as possible.
>
> Some logistical things, there is not adequate parking for every
> volunteer to drive and park at the Computing Center. We
> would ask that
> volunteers either carpool (4 or more to a vehicle) or take the BT.
> Initial cabling will be done with the power cables beginning this
> coming Friday, August 1st. We have yet to set a start time,
> but I want
> to get an idea of who is available to assist on that day as well as
> availability for the weekend and early next week. I will forward the
> time to arrive as soon as I nail it down...should be tomorrow. The
> power cabling represents one third of the overall cabling that will
> need to be done.
>
> I need to get a list of those who will be assisting to the facilities
> people at the ISB as soon as possible. This list will be
> used to allow
> volunteers access to the building and the machine room. If
> you're not
> on the list you will not be allowed in! We want to have at
> least 30 to
> 40 people working at any one time on the project. If you know others
> that will be interested in assisting, please have them email me so I
> can get them on the roster.
>
> Thank you for your willingness to participate, and please respond and
> have others respond with "Terascale" in the subject so I can
> filter the
> volunteers properly.
>
> Thanks.
(E-mail signature removed)

I am peeing myself...

[ErnstKompressor]

I never thought the mac-madlib-trollbot-AC post had the power to do more than infuriate...

I stand corrected.

Bravo.

The FASTEST CLUSTER is a 21364 Alpha

[Anonymous Coward]

You can't beat a true 64bit platform, from the beginning to the present; Alpha has always been 64bit and the leader of high-performance computing.

Supercomputer Research is Vital

[reporter]

Supercomputer research is an important area of research that fits well within the academic environment. Indeed, a generation of American (not foreign) computer engineers were trained at universities like Virginia Tech by doing research on a large-scale supercomputer project. Please read "American Supremacy in Supercomputers".

Re:Supercomputer Research is Vital

[Anonymous Coward]

Linking to your own comments?

You're a sad, sad little man.

Think Secret's Record

[Nick dePlume]

As Zack pointed out, iWalk was not a Think Secret report; in fact, we debunked it. For WWDC, we reported that Apple would announce 64-bit Power Macs as well as a videoconferencing camera that we said would be called "iSight," -- I think we're in the clear there. iWorks? I maintain that it is still a future Apple release. As for 12-inch and 17-inch PowerBooks, while we raised the possibility of a release that week, we specifically said we couldn't confirm the delivery date: "It's unclear when Apple plans to announce the upgrades..."

Bottom line? Like any other news organization, Think Secret has occasional misses. But those misses don't appear to include any of the items mentioned here. I think our record speaks for itself.

Nick dePlume
Publisher and Editor in Chief
Think Secret

Re:Think Secret's Record

[Ninja Programmer]

Bottom line? Like any other news organization, Think Secret has occasional misses.

Most news organizations don't report rumors as anything other than rumors and write retractions when they get it wrong. How is your record there?

Re:Think Secret's Record

[Nick dePlume]

I'd argue that it's solid in that regard as well.

I believe we're honest and up-front about the reliability of what we report. If something has not been confirmed by multiple reliable sources, we say so. If something is mere rumor, we also say so, but most of the time, we would rather not report it if we can't confirm it.

As for mistakes, after Macworld Expo, for example, we typically take a look at the announcements and compare them to our pre-expo reporting to see what, if anything, we got wrong. We don't hide anything.

Nick dePlume
Publisher and Editor in Chief
Think Secret

Re:Think Secret's Record

[inkswamp]

I think our record speaks for itself.

Your site won me over when you were the first and only site to announce unequivocably that iTools would become the $99/yr. .Mac service. Everyone posting at every other Mac forum decried it as impossible or unlikely or just plain wrong; it's funny to think back to all the explanations and discussions of why it simply could not happen, never, ever, end of story and how Apple would lose customers and go out of business and it would be the end of civilization as we know it so that rumor could not be right. Even I thought you guys had gone off the deep end.

Needless to say....

Talk about a ton of desktops in a server room

[Kalak]

For the ones who are questioning this existence, the order is shipping, the racks (a ton of them) are there in the main Computing Center server room. First they required all servers to be moved innto racks. Then they started moving servers around, including removing the Petaplex. The power has been upgraded in the server room (the UPS backup generator actually). This caused a morning of basically all the important servers on campus having to go down for one day in the summer - I hated waking up to go switch off machines for that one. The AC has been upgraded to accomidate the huge amount of heat to be put out. It was't until I heard about the cluster that all the chages in the Machine Room made sense. Now they're recruiting help to do the grunt work of putting all the machines in the racks.

The stated objective was to be on the next 500 list. Dell and HP were considered, but they couldn't fill the order in time (possibly as they have made announcements of other large clusters recently) and Apple promised delivery after someone leaked the story of the cluster meetign with Dell and HP to Apple and Apple jumped at the chance.

Basically, the story is not a rumor from the point of view of the geeks on campus who have been effected by the preperations. I'll probably post the /. link to the campus geek list (If someone hasn't beaten me to it).

I'm disapointed about this being only on the Apple section of /. since a cluster this size is noteworthy of the frontpage. (Rumor - and this is rumor sice I haven't goe to direct sources on this - is that it will not be running OS X, and probably BlackLab or YellowDog or SuSE.)

Re:Talk about a ton of desktops in a server room

[Anonymous Coward]

Rumor - and this is rumor sice I haven't goe to direct sources on this - is that it will not be running OS X, and probably BlackLab or YellowDog or SuSE.

Oh, gosh! That's like USD 768900 in license fees to SCO!

(Posting this as AC, since this is getting really old.)

Re:Talk about a ton of desktops in a server room

[valdis]

Actually, this summer's outage for the new diesel backup generator was something else entirely - that was merely replacing two older natural-gas fired generators that were no longer sufficient to fully back up all the existing hardware. That install has been in the planning stages for a long time, and was needed for current operations.

Do the math - the new generator is rated at 600kva and is already carrying several hundred machines (including a very power-hungry Sun E10K and a number of E6K-class machines). There's not enough capacity on that generator for 1,100 more systems.

(And I just wanted to say "This is All Kevin's Fault" - except for a few unrelated parts we blame on Randy ;)

Re:Talk about a ton of desktops in a server room

[Wesley Felter]

Does Linux run on the G5 yet? I've seen no mention of it on linuxppc64-dev or linux-kernel.

Re:Talk about a ton of desktops in a server room

[Johnny Mnemonic]

I don't believe so. At least Yellow Dog Linux hasn't released a distro for the G5 yet, and promise to when they are able.

Re:Talk about a ton of desktops in a server room

[Kalak]

Make sure the G5 vigilanties DDoS Kevin instead of pulling an SCO style DoS. We could make more jokes about Randy being used to DoS, but this is /. - we don't want to appear mean do we?

Re:Talk about a ton of desktops in a server room

[soundofthemoon]

Actually, it's more like 20 metric tons. That's 55 cubic meters of PowerMac!

Re:Talk about a ton of desktops in a server room

[N2UX]

including a very power-hungry Sun E10K

C'mon. The E10K isn't power hungry. Its the multi-terrabytes of disk attached to it. Now, if the Terascale facility were built of a cluster of 1100 E10Ks - THAT would be power hungry!

Re:Talk about a ton of desktops in a server room

[G4scott]

If it's going to be running linux, how will the fans be controlled? Or will they be on full speed all the time? I ask because I heard the fans were controlled by the OS.

I was also wondering, why PowerMacs, and not xServes, but then the answer hit me... The time constraints. I'm sure that if they wanted a super computer cluster in the next 6 months, Apple would've been ready with G5 xServes, but hey, 1100 top of the line G5's will still kick some major @$$...

Re:Talk about a ton of desktops in a server room

[Johnny Mnemonic]

There's not enough capacity on that generator for 1,100 more systems.

Well, I guess you had better find some capacity. C|Net just confirmed this story. If these machines aren't going where you thought they were, what other provisions have to be made? I'm curious to know, and it sounds like you have exposure to the facilities in question.

Re:Talk about a ton of desktops in a server room

[valdis]

Exposure to the facilities? All I'll say there is that I know of at least 4 people who have posted to this story that could throw a paper airplane from their cubicle and have a reasonable chance of landing it on the blueprints. ;)

I said there wasn't capacity on that generator. I didn't say anything about the existence of other generators - and there's a distinction between pulling copper to get power grid capacity to the cluster and having emergency power backup for same. That diesel is for emergency backup, not local power generation - we've got a nice coal-fired beast on campus for that.

In addition, you have to remember that the cluster most likely has different backup power requirements than our production systems. Our production systems (the central mail server, the large database machines, etc) really need to stay up - we're talking about things that if the power substation near the airport goes out, the whole campus is screwed till it comes back online, if we don't have backup power.

Look at how many services were out when we did the cut-over - everything was affected, from E-mail to the library catalog. That's what emergency power is for - so you don't have an event like that without the multi-week notice that the install had.

On the other hand, does a compute cluster really need that level of power backup, or can we significantly trim the budget by merely having a really good power conditioner to make sure there's nice clean non-ripply power, and enough battery backup to allow the cluster to do a suspend-to-disk and poweroff cleanly? Yes, if we take a power hit in that scenario, some researchers have to wait for their results - but the business of the university as a whole isn't disrupted.

Or possibly the plan is to have the conditioner and batteries for now, and have someplace to cable in a backup diesel generator in the future - remember that 1,100 G5s, a bunch of Infiniband switches are a chunk of change - and then you get the expense of the power/cooling work. Suddenly you start thinking about how to put the non-essential costs into the next fiscal year. ;)

(The above is not intended as an actual statement of the actual design or plans, just an illustrative discussion of the fact that there are a lot of tradeoffs that need to be made when planning and designing large-scale installations of any sort).

Re:Talk about a ton of desktops in a server room

[Anonymous Coward]

Wow you seem to know a lot about this and giving out a lot of info too! Other posts suggested this was supposed to be really hush hush!--------guess you don't play by the rules.....or there is more to tell.

I Think He's Talking About "The Blue Devil"

[Farley Mullet]

Maybe you were just making a joke....I had no idea. :)

Yep, he was. Y'see, you Dookies usually like to talk smack about how good your teams are, especially the basketball team. So you might be "in" on the computer stuff, but you seem to be "out" on the campus-wide obsession stuff.

Bonus Quiz: How many Coach K-coached players have gone on to win NBA titles?

Re:I Think He's Talking About "The Blue Devil"

[leviramsey]

I know! I know! Exactly one!

And it only took about 20 years!

Talk of it all over campus? Mouse "balls".

[Anonymous Coward]

" Yeah, chicks dig massive...computers.

No wait, no they don't!"

And yet most male geeks see no problem with playing with their mouse all day long.

For the non-hokies...

[Spleener12]

For those of you who are non-hokies, the Math Emporium is a slightly off-campus location that people go to to take their math tests/quizzes (among other things, too, but that's why I have to go there.) It's in, of all places, a mall (a very small one, at that.) I just went there the first time a few hours ago, and my first thought upon entering there was "Wow. I didn't know Apple PRODUCED that many of those iMacs..."

Imagine, if you will, a room about the size that a mall deparment store might be if they took out all the walls separating the individual departments. And in there, there are HUNDREDS of iMacs (the 700 mHz ones with the LCD screens- I checked the clockspeed myself on the one I used.) And along (at least) one wall, there were god-only-knows how many more G4 towers and monitors. It was an Apple geek's nerdvana.

At first, I figured it was just because the math department liked Macs better. But now that I think about it, all that people did on these computers was take tests and quizzes through a web browser. You really don't need that much power to do that, and the iMacs were the cheapest, easiest-to-set-up (remember, there were HUNDREDS of these fuckers in there) option they had.

1100 Nodes....

[Anonymous Coward]

And still only one mouse button

As a VT student, the answer seems quite clear

[jdog1016]

It's because we can.

Duke Sucks

[Anonymous Coward]

Sorry I have been reading too much Fark.

How To Pick An Arbitrary Constant

[titzandkunt]

Here's how to pull an arbitrary constant out of your ass, without getting pulled up in that pesky code review. Your colleagues will think they should know the significance of your constant, and will keep their mouths shut...

2^n

(2^n)-1

Anything in hex, eg. 0xf67d34ee, but if you're going to pull this stunt, make sure you #define DEBUG_MEM_VAL 0xdeadbeef as well, so they think you know what you're doing.

Never, never, never use a round decimal value. T&K.

MOD PARENT UP

[Anonymous Coward]

only +2? The RDF is working nicely today.

a Hokie alumn speaks

[rtphokie]

Virginia Tech placed the dual-2GHz G5 order shortly after the G5 was announced. Multiple sources said Virginia Tech has ordered 1100 units

Ummmm, Virginia Tech (like many institutions of higher learning, especially in Virginia) is counting paperclips. Tuition is skyrocketing and state financial support is drying up.

There has been talk of going private and actually coming out ahead because the amount of money coming in from the state is fast dropping to the point where all the people required to maintained the state mandated beurocacies costs more.

And where did the $$$ for these 1100 G5's come from?

Re:a Hokie alumn speaks

[pyr0]

Not much to add to what you said...but I was wondering the same thing. I *just* started here at VT (getting my PhD), and from everything I've heard the budget crisis is very bad. Lots of classes and labs have been cut down on the number of available slots...or even completely cut period.

Re:a Hokie alumn speaks

[Anonymous Coward]

As was mentioned in another post, alot of the money for this project came from research grants.

Re:a Hokie alumn speaks

[wayward_son]

Virginia isn't the only state that is having problems with funding.

Clemson only gets 25% of its funding from the state.

The Medical Univ of SC only gets 4%! They might as well go private, especially after the governor said that South Carolina doesn't need two state medical school.

(For those not from the Palmetto State:
Medical Univ. of SC = Charleston
Univ of SC School of Medicine = Columbia
It really is that confusing.)

Re:a Hokie alumn speaks

[stingerman101]

If the money granted was earmarked for a specific project their is nothing the university can do to transfer it to another need. It is either they use it or lose it. Fortunately these grants usually include monies for maintenance of the systems and the staff required, which means that personnel can be put under this new grant area and alleviate costs from another area. So in the end, grants like this help other areas of the school and will attract and create other sources of revenue.

Re:a Hokie alumn speaks

[Dros68]

The letter announcing the cluster says they will be able to attract new federal grant support with the new capabilities. In the sciences, grant money includes up to 50% overhead money that goes to pay off creating and maintaining infrastructure. Most schools recognize that having high-quality research programs brings in enough money to make those units self-supporting or even help subsidize other aspects of the university. So for a few million dollars, they'll hope to bring in a few million a year in grants, as well as attract high-quality researchers who will bring in more grant money. Compare that to the usual revenue gamble at universities-- spend 100-200 million on new football stadiums and facilities to vie for bowl game money that will hopefully pay off year-to-year costs. Usually, the stadium expansions get approved when a team hits a win streak, and by the time it is built other teams are better and they are left with half-full stadiums.

quote "but there's talk of it all over the campus"

[Anonymous Coward]

Thats five geeks talking over coffee.

Redeeming Value

[Anonymous Coward]

Go to Radford University parties just a few miles away. Drink lots, then 'Go Ugly Early'. You libido will find redeeming value.

Real question

[McAddress]

thats cool, but will it run Linux?

1% of G5 orders

[Viadd]

Just for perspective, there have been over 100,000 G5s ordered, so this cluster is about one percent of the backlog. In other words, assuming that Apple ships all pending orders in about a month, the G5 I ordered will be delayed by about 8 hours.

And 8 hours@12.4GFlops...damn you Virginia Tech, you owe me a third of a quadrilion floating point multiplies!

Keep that __ Browser

[rixstep]

Hey maybe IE for Mac will be faster now?

Re:Keep that __ Browser

[scrod]

I think all those usleep() calls will still hold it up, unfortunately.

1100 machines is ~1.1% of the placed orders

[vnv]

According to Apple, there were "over 100,000" pre-orders for the G5. Now this includes single processor models, but the university's alleged order of 1100 machines is not going to make a big impact on everyone else.

Besides, the real reason that Apple's machines are late is case defects and AGP problems, amongst other issues that Apple has not been forthright about. At the keynote an honest Apple employee told me the machines wouldn't ship until October as there were many little problems and I should wait for the January refresh so I don't get a flaky machine.

And one has to wonder why anyone building a cluster would build it using desktop machines and not use the forthcoming G5 rackmount machines from Apple and IBM... which is supposed to include a quad-processor from IBM.

Re:1100 machines is ~1.1% of the placed orders

[stingerman101]

Delivery time was obviously the key issue. The grant probably depended on it happening within a time frame, and Apple jumped to meet it.

Re:1100 machines is ~1.1% of the placed orders

[vnv]

That still doesn't change the fact that Apple apparently delayed tens of thousands of G5 shipments for a month, making other companies and individuals all wait just to please this one university. If I had lost my place in the queue, I wouldn't be happy having to wait longer for a machine I likely ordered three months ago.

Now what maybe true is that because these machines are going into a cluster, they don't care about cosmetic problems with a patched case door and they don't care if AGP 8X doesn't work right for high-performance 3D cards. So Apple could be dumping 1100 "good for clusters" machines on the university while waiting for an inline revision of the first batch of G5's. That makes more sense.

And you know... I'm quite glad I took an honest Apple employee's advice and don't have to worry about this stuff. Come January (Feburary is what I put on the calendar), I'll have a G5 with some of the bugs worked out and for less money.

G5 Vs. Itanium2 and Opteron: Some perspectives

[Anonymous Coward]

1. The PPC970 draws from the Power4 lineage, which I have used for a long time. The PPC970 has 2 double precision FPUs, each capable of fused multiply add instructions leading to 4 flops/cycle/processor (2 units*2flops/cycle). This is identical to the Itanium2 FPU microarchitecture. The Opteron on the other hand can only do 2 double precision flops/cycle, which makes it only half as powerful on matrix heavy scientific computations, when compared to the PPC970 or the Itanium 2. The PPC970 should really be compared in FP terms to the Itanium2 at 1/10th of the cost, and at 2GHz it is clocked higher than the top-end 1.5GHz Itanium2 Madison. Moral of the story, read thy arstechnica. 2. The standard benchmarking process (LINPACK) only uses double precision FP. If this rumor is true, then this machine is capable of an Rpeak (LINPACK) of 17.6 Teraflop, which those of you who follow top500 will realize is quite substantial. 3. If they are really using Infiniband, this should be a nice machine. Infiniband provides 10 Gbps (20 Gbps full duplex) of bandwidth, which is much faster than either Myrinet or Quadrics. Also Infiniband latency is 10us and the benchmarking process is bandwidth not latency sensitive. On the other, this stuff is really expensive. If all of this is true, this would be a major engineering endeavor. Also, it is probably cheap. However, all in all, this could well just be a rumor (come on it is thinksecret - remember iWorks). If not, this should be a fairly substantial machine.

Re:G5 Vs. Itanium2 and Opteron: Some perspectives

[stingerman101]

From manufacturer's specs on the Infiniband cards being ordered for the Virginia tech project: Tests have proven that these HCAs provide up to 8 times the bandwidth of gigabit Ethernet. Additionally, tests conducted by Ohio State University have shown over 820 MBytes of delivered MPI bandwidth which is up to three times the bandwidth of existing proprietary high performance computing clusters interconnect technologies. Zero Byte latencies achieved are as low as 7.5 usec and large message latencies are as low as 1/3 of proprietary HPCC interconnects.

Re:G5 Vs. Itanium2 and Opteron: Some perspectives

[stingerman101]

http://www.mellanox.com/products/shared/HCA_PO_050 _all.pdf

Re:G5 Vs. Itanium2 and Opteron: Some perspectives

[J0ey4]

Mod parent up to +5, he is the only one to make a comment on this page that truly understands why this will be a monstrous machine.....

SGI Origin 3000, 1024 processors...

[green pizza]

It's just too darn hard to make a shared memory computer with 1000's of processors. So the common architecture is to make a cluster of smaller shared memory machines.

It's hard, but not too hard or impossible. The Silicon Graphics Origin 3000 supports 512 processors in a single image system with the stock IRIX kernel and 1024 processors with the "XXL" kernel.

Rumor has it Origin 4000 will support 2048 processors, as will Altix once SGI has done some major work with their kernel patches. (Altix is currently limited to 64 processors per system image).

Why G5's? I helped set up the racks

[Anonymous Coward]

They went with the G5's because they were the cheapest 64 bit solution and because they would use less power and generate less heat than alternative systems. That is the whole of it.

They are having someone write infinaban drivers for OS X just for this cluster.

I look forward to helping install 4GB of ram + the infinaban cards in each of these bad boys.

It is great having connections!

What OS?

[Anonymous Coward]

It will be OS X. They were looking for Darwin kernel hackers to write Infinaban Drivers and I know the people who will be using the system. Hopefully I will get to use it this fall for my parallel computation class projects.

They paid Educational Pricing

[Anonymous Coward]

They paid educational pricing straight up with no bargain from Apple. Infact, Apple was hesitant/scared at the prospect of such new/untested hardware being used in such a high-profile project. I believe they will be running the stock OS, ie: not OS X server as these are not server machines.

It is a great time to be a hokie in CS4234 aka. Parallel Computation!

Congratulations!

[Farley Mullet]

K came along in '80-'81, and Ferry (finally) got a title in 2003, so that's 22 years. And what's a Dookie doing at UMass? Shouldn't you be representing for your Minutemen? Sure, they've kinda sucked since Coach Cal cut and run, but c'mon dude.

Re:Congratulations!

[leviramsey]

I kind of inherited it from a sister who's a Duke fan. Being a UMass basketball fan is just pitiful and I won't stoop that low.

Bragging rights...

[ca1v1n]

As long as they put it on Legion, U.Va. will get to maintain state bragging rights. I don't keep up with the football rivalry, but this is much cooler anyway.

Rock and Roll and Basketball

[Farley Mullet]

See, the thing is that Duke doesn't have a great song written about it like the Pixies' UMass, which was what made me a Mass fan, back in the day -- and the early nineties were when Coach Cal and Marcus Gumby and that wicked backcourt of Padilla and Travieso were tearing up the nation (plus comically named forward Dana Dingle!). But yeah, ever since then, it's been lean years for Mass basketball.

w00t VT!!!!

[Anonymous Coward]

I go to VT, and while I'm dissapointed that it's a shitload of G5's... It's still pretty sshweet.

VT is the freakin best man. We've got the tech, we've gots the women (and radford chicks)

On a side note, I'm fucking wasted right now... WHAT THE FUCK ARE YOU GOD DAMN SLASHDOTTERS DOIN SATURDASLY NIGHTS READIN THIS SHIT???!!!

You should be out partying, or havin fun, or fuckin, or something. I may be a fucking geek, but even I know when to get wasted!!! The only reason i'm postin is cause me freind IMed me about it while I was out partying, and told me to comment, cause I'm from VT.

HERE'S MY FREAKIN POST YOU LOSER BIZNITch!!! GO DRINK SOME BOOZE ON A SAT NIGHT!!!!

w00t!!!

AND WHOEVER WE BE PLAYIN TOMORROW, YOU'S GOIN DOWN!!!!

GO HOKIES!!!

*end*
drunken man rant
(and yes, i am drunk) :)

G5 Ship dates have just changed to 10-15 days..

[Corpus_Callosum]

Just for perspective, there have been over 100,000 G5s ordered, so this cluster is about one percent of the backlog. In other words, assuming that Apple ships all pending orders in about a month, the G5 I ordered will be delayed by about 8 hours.

First, let me say that I think it is good for Apple and therefore good for Apple users that they are supporting the VT supercomputer project by bumping their priority. This will be the first time that Apple computer will be able to claim anything like this and it should mark an important turning point in the minds of corporate and scientific buyers regarding the strength of Apple's platform. And I agree completely, 1100 machines is not going to delay anyone's order by much (or at all).

Speaking of delays: It may be a mistake at the Apple Store site, or hopefully (maybe) it is by design! - I went through the process of purchasing a dual 2ghz (stock) G5 machine to look at the shipping date and guess what? Estimated ship: 10-15 bus days. Go check it out for yourself... Is this real, or a Labor day, labor problem?

Linux is #3 on Top 500

[jmichaelg]

If you look at the top500 list, you see Lawrence Livermore's Linux cluster is at 3rd place. It's an 1100 node cluster of 2.4 GHZ P4's. Looks like Apple is a bit late to the party if they're only shooting for #5.

Re:Linux is #3 on Top 500

[Anonymous Coward]

"Top 5"

Re:Linux is #3 on Top 500

[stingerman101]

Yesterday's news, we're talking about the next list update. With the XEON P4 at number 3, it should be easy for this new G5 cluster to be in the top 5.

NERDS!

[Anonymous Coward]

Unless the entire campus consists of five dateless wonders, I don't believe the entire campus is talking about it.

Re:NERDS!

[Anonymous Coward]

I don't believe the entire campus is talking about it.

That's not what it says dumb*ss, it says:

but there's talk of it all over the campus

Clearly that means that the nerds are just walking around.

That's not too much of a stretch, is it?
(Usually they tend to congregate at the CIS building, but I'm not sure how spread out the IT department is there. So it's possible there are labs sprinkled around the campus that would substantiate this claim.)

apple marketing fagboys

[Anonymous Coward]

is what it is. please stop with your spin machine - fagboys.

Easy to find out

[Escoutaire]

Just keep an eye on the skips (dumpsters) where the computing facility dumps it's rubbish. The packaging from the new machines will be easy to spot.

Escoutaire

Re:[insert chip/os vendor here] marketing fagboys

[Anonymous Coward]

is what it is. please stop with your spin machine - fagboys.

Damn, I guess we're all fags now... Any other bright ideas, genius?

This is getting really old

[nusuth]

It is very easy, Altivec can do something SSE(2) can't. If both ISAs can vectorize the given operation, SSE2 is just as good. With higher frequency, SSE2 wins. When you need to do something that can be done with altivec but can't be done with SSE(2) unit (inter-vector permutations are one example, see arstechnica.com for more examples) altivec is way faster.

Also, I don't know how much PPC970s costs but considering how much dual G5s from Apple are expected to cost, PPC970 must be very competitevly priced.

7/16/03

[pjdepasq]

Someone here (me) scooped this on /. back in July. Welcome to the party.

Perhaps they are simply loading Darwin?

[emil]

At which point the software costs drop to zero.

Apple better watch out!

[Anonymous Coward]

Those guys at Virginia Tech would sell their own mother for a couple of bucks! Apple better watch out and hope that HP or IBM don't sweep in and offer to do it for a few bucks cheaper!]

Summary: VT is a collection of filthy whores!

MOD PARENT DOWN

[Anonymous Coward]

Yet another AC telling moderators how to mod another AC. Perhaps they were busy following up the posted link (useless) and researching the accuracy of the claimed 9 hour up-time. The obvious issue is the validity of any results, not if Word is likely to lunch in the middle of an auto-save, (the answer to which is always yes).

AMD x86/64

[mslinux]

Anyone know how much less expensive a cluster of AMD x86/64 machines would have been? How would the performance differ? If x86/64 is less expensive and performs on par (equal to or slightly less than) with a G5 cluster, then why not use that instead???

Apple must have given a *big* discount...

Re:AMD x86/64

[stingerman101]

Well the only AMD alternative is the 2GHz Opteron whose processor alone costs about 900/processor. You won't be able to get a dual 2GHz Opteron for less than Apple's retail dual 2GHz G5. Too, you have to make sure PCI-X slots are there for the interconnect cards they are using. But as was stated elsewhere, the G5's can process twice the scalar FP instructions of the Opteron or the Itanium per cycle. The University as also relies on vector math and the G5's have a significantly better SIMD unit.

Nah...

[Anonymous Coward]

Nah... 10 GB ethernet is capable of 1280 MB/Sec or 1.2 gb/sec without overhead.. overhead will probably kick that below 1 gb/sec by a bit... but thats still around the speed of pc100/pc133 memory. While pc133 is not the fastest memory available today, it's not horribly slow either.

You could double, triple, or quadruple up the ethernet with multiple nics, etc... and I think one could hit the speed of memory bandwidth over ethernet easily.

The latency would probably be an issue though. But bandwidth itself - nah it oughta be OK.

I'd think...

RC5 Speed

[charnov]

The main reason that RC5 goes so fast using AltiVec is that it includes a Vector Rotate instruction while every other SIMD engine does not. A Vector rotate is a very rare instruction (and trivial to add to SSE, etc.) that RC5 makes extremely heavy use of.

That is the ONLY reason why RC5 goes that fast on a G4/G5.

Imagine this!!

[Anonymous Coward]

Imagine a beowulf-cluster of... eh.. oh, wait...

But, what will they *do* with it?

[chiph]

Sure, having a top-5 supercomputer is cool, and the bragging rights will garner the school some additional funding and scholarships, but at the end of the day after the benchmarks are run and the empty Jolt(tm) cans are recycled, what will they *do* with it?

Chip H.

probably can't get

[waspleg]

halflife playable in software mode either

i think we need quantum computing faster

and where are my solid state harddrives and flying cars and genetically engineered super model sex slaves while you're at it...

aaditya@member.fsf.org

[Anonymous Coward]

aaditya@member.fsf.org

Correction:1100 Dual G5's = 2200 CPUs

[Anonymous Coward]

Yes, a beowful....of 2200 CPUs, probably running Suse Linux Enterprise (64bit OS)

This is all so totally ridiculous!

[sylas]

Okay, so I'm browsing a tech news page and come across al ink to this thread . It piques my interest and I start reading some of the posts. Then it occurs to me how totally stupid this whole thing is. People on this thread (and plenty more like it on this and a multitude of other sites) are arguing, debating, waxing poetic, philosophizing, etc., etc., about man-made, highly engineered chunks of crystal. How f'n stupid is that? I was a "car guy" at one time and we could bitch for hours about the relative merits of the Chevy vs. Ford thing. But I don't recall there being this huge pissing match over how the engines worked. If I'm not mistaken, there are no little elves that sneak into the labs at IBM and Intel at nigh and weave magical, unknown architecture into these chips. It's all plotted out very methodically by humans and computers (designed by the humans) running programs (also designed by the humans) and there is not AI or unknown natural intelligence involved. It's all there on the map. My main point is this: in the grand scheme of things, who cares! My god, people, put those big f'n brains to work on something really mysterious like the natural sciences, or something creative. Put some of that wasted effort into trying to figure out how your own thinker works. Now that's interesting and full of mysteries and possible explanations. This is not meant to be a troll post. Please don't take it as such. Just MHO.

Apple Parity Checking

[vought]

Apple did offer a parity-checking SIMM option on the IIci, from 1989-92, I believe.

Re:Why G5's? I helped set up the racks

[jcr]

Four gigs? Why aren't they maxing them out?

-jcr

Maybe Acrobat will launch faster

[Toe, The]

I'll bet on this cluster, launching Acrobat Reader will only take like 30 seconds.

(As far as I can tell, Adobe has come up with some sort of reverse Moore's Law, where each new version of Reader takes twice as long to start up... at least on Macs.)

Re:Maybe Acrobat will launch faster

[shamino0]

As far as I can tell, Adobe has come up with some sort of reverse Moore's Law, where each new version of Reader takes twice as long to start up... at least on Macs.

Windows too. When I upgraded from 5.1 to 6, the startup time became unbearably slow on my P2/300 system. It's only barely acceptable on my P3/667 system. (Both running Win2K with 256M RAM.)

Now public

[Anonymous Coward]

The University has now gone public with this

http://www.vt.edu/news/showitem.php?id=106260174 6

CNET Article

[beetle496]

Apple shooting for supercomputer heights:

Apple Computer has landed a major customer for its Power Mac G5, with Virginia Polytechnic Institute and State University confirming Tuesday that it will use 1,100 of the machines as part of a supercomputer cluster now under construction.

It doesn't matter what the OS is.

[gaudior]

Apple is a HARDWARE company. They make a really cool Desktop OS, and a really cool General Purpose Server OS. They do not necessarily make the best OS for a dedicated number-crunching cluster. You don't need all that Aqua/Quartz lickableness if all you are running is massive Fortran matrix apps.

They will distribute to other departments

[Smurfman]

When they upgrade to newer, faster machines, they can distribute what they want to other departments!

Wow

[gsparrow]

How much does that cost?

Re:Why G5's? I helped set up the racks

[shamino0]

Four gigs? Why aren't they maxing them out?

Availability? I just did a quick search on PriceWatch and nobody is selling 1G PC3200 modules. All of the listings for that capacity are actually selling "kits" consisting of a matched pair of 512M modules.

The fastest memory listed there that comes in 1G sizes is PC2700. The dual-2GHz G5 systems need PC3200.

Of course, to be fair, I did notice that Apple seems to have access to 1G PC3200 modules. Systems ordered from the Apple Store with 2G or more RAM have them installed.

The real reason they aren't maxing them out probably has a lot to do with the software they plan on running. Depending on the application and the data set, they may not gain anything in the move from 4G to 8G. And it greatly increases the cost (using Apple Store pricing, 8G costs $2600 more than 4G.)